News & Updates
The latest news and updates from companies in the WLTH portfolio.
Anthropic Achieves Profit, $16.8B Q3 Revenue
Aims for $2T Valuation in IPO, Plans AI and Data Center Investments Anthropic, a U.S. artificial intelligence (AI) company pursuing an initial public offering (IPO), recorded Q3 revenue of 16.8 billion dollars, approximately 22.5 trillion Korean won, a 45% increase from the previous quarter. According to Reuters on the 14th (local time), Anthropic disclosed these results at a preliminary investor presentation for the IPO held at its headquarters in San Francisco, U.S. on the same day. The company's revenue grew rapidly from 4.7 billion dollars in Q1 this year to 11.6 billion dollars in Q2 and 16.8 billion dollars in Q3. Notably, the company also achieved a profit on an adjusted operating income basis in Q3 following Q2. This contrasts with last year, when despite recording annual revenue of 4.6 billion dollars, the company posted an operating loss of 8 billion dollars. The disclosure of these results is seen as a move to demonstrate growth potential and profitability to investors ahead of the IPO. Anthropic aims to be valued at over 2 trillion dollars, approximately 2,700 trillion Korean won, during the IPO process. The secured funds are planned to be used for large-scale infrastructure investments, including AI model development and data centers. Anthropic had initially planned to list this month, but with the schedule delayed, the possibility of listing after the U.S. midterm elections in November has grown. OpenAI, which also pursued a listing within the year, has postponed its listing date to next year.
Bank of America CEO reveals how US' second-largest bank had to hire thousands and spend millions of dollars to 'contain' damage done by Anthropic's AI Agent Mythos
America's second-largest bank, Bank of America (BoFA), may have just told everyone how dangerous AI agents can be for companies. According to Bank of America CEO Brian Moynihan, the bank has enlisted an army of external contractors and deployed thousands of tech workers to fortify its code against emerging cybersecurity threats posed by frontier artificial intelligence. CEO Brian Moynihan revealed the same in an interview with CNBC. The defensive overhaul was reportedly triggered after advanced AI Agents, including Anthropic's Mythos -- demonstrated a heightened capacity to pinpoint and exploit longstanding software vulnerabilities across bank's internal and vendor networks. "That risk is real risk, and that's why when the frontier-type models came out... a lot of us sat down and said, 'How do we figure out how to get through this quickly?'" Moynihan told CNBC's Leslie Picker following the release of the bank's third-quarter earnings.To deal with an immediate threat, the bank hired 1,000 third-party contractors during the second quarter. Those specialists joined 3,000 existing internal staff dedicated to software maintenance and security, Moynihan said, and they remain actively engaged in remediation efforts.The emergency response meant a substantial cost for the bank. Moynihan told analysts that the bank incurred a "couple hundred million plus" in expenses during the third quarter alone, tied to the software patching campaign alongside anti-money laundering regulatory compliance measures. Moynihan's remarks offer an unusually transparent glimpse into the tangible dollar and labor costs that major US financial institutions face as they confront the unintended fallout of rapid AI innovation. Mythos was distributed in April to a select cohort of banks, technology firms, and government agencies, setting off an urgent defensive scramble across corporate America to remediate decades-old coding gaps before malicious actors could take advantage of similar frontier tooling."We've got to stay ahead of them, and we've got to be mindful that the people doing the software development are the people trying to sell us the services," Moynihan said.The spending disclosures comes against a challenging market backdrop, with Bank of America shares sliding as much as 8%. Addressing investor concerns over escalating overhead, Moynihan emphasized that the surge in cybersecurity contracting would be temporary. "Once you patch, you don't have to patch as much... the pipeline moves through," Moynihan said, noting that remediation expenses are projected to taper off once core vulnerabilities are sealed. 'Bank of America is replacing only 75% of employees who leave' Even as the bank absorbs short-term defensive costs, it is actively leaning on AI to drive broader operating efficiencies. Bank of America currently operates roughly 130 internal AI initiatives. The bank expects annual cost savings generated by those systems to jump from $300 million this year to $700 million by 2027. Over the past year, the lender reduced its overall workforce by roughly 3,400 positions, a drawdown that included shedding about 5% of its management ranks. Speaking at an event, Brian said that BofA has an employee attrition rate ranging from 8% to 8.5%. Based on current headcount, the implication is that around 1,400 people leave each month, at least. When they do, he said that the Bank only replaces 1,000 of them. The other 300 jobs must be 'owned' he said. More prosaically, they must be done by existing staff.Fortunately, AI is around to help with this. "In the end of day, our teammates have got to use AI to make our company better and make our clients a better experience," declared Moynihan. BofA now has "150 use cases implemented" and is deploying AI into "more work," he said too. This work includes audit, legal and other areas. And BofA's 18,000 coders have "saved" 10-15%, which is being redeployed. Bank of America fined $1.2 billion The US government has imposed a $1.2 billion fine on America's second-biggest bank. The amount is the third-largest in the department's history. "When a bank of this size fails to meet its anti-money laundering and sanctions obligations, the consequences extend far beyond its own walls, exposing the entire U.S. financial system to significant illicit finance risks," Treasury Secretary Scott Bessent said in a statement. "Today's actions send a clear message that even the biggest financial institutions are not above the law." In a statement, Treasury said that its Financial Crimes Enforcement Network (FinCEN) assessed the $1.2 billion penalty for Bank of America's "failure to mitigate and report priority national security risks, including Chinese Money Laundering Networks and illicit finance along the southwest border.""Financial crime compliance is serious, and we are responsible for getting it right,'' Bank of America said in a statement. It added, "We have invested billions of dollars in technology and dedicated thousands of employees to combat financial crime, and our team is far along with and committed to completing the work required to address the issues raised by our regulators."
Anthropic Q3 revenue surges 45% to $16.8 billion as AI startup gears up for $2 trillion IPO
Anthropic reported a 45% increase in revenue for the third quarter, reaching $16.8 billion. The company's revenue for the first and second quarters was $4.7 billion and $11.6 billion, respectively. Anthropic aims for a valuation exceeding $2 trillion, emphasizing substantial revenue growth despite significant operational losses. Founded by former OpenAI executives, the company focuses on AI safety amid growing competition. Anthropic's third-quarter revenue rose 45% to $16.8 billion from the previous quarter, according to an investor who attended a pre-IPO meeting with Anthropic leadership in San Francisco on Wednesday. Revenue in the first quarter was $4.7 billion, and $11.6 billion in the second quarter, according to the person who attended the investor day meant to familiarize prospective Anthropic investors before its highly anticipated IPO. The company was profitable on an adjusted basis in the third quarter, Anthropic told the group, the source said. Also Read: Anthropic targets more than $2 trillion valuation in closely watched IPO despite $42 billion 2025 loss Anthropic did not immediately respond to an email seeking comment. The IPO is expected to precede a public listing by rival OpenAI, making Anthropic an early test of investor appetite for heavily unprofitable AI developers pursuing massive scale. Its public-market debut is likely to be pushed until after the November US midterm elections, Reuters has reported. Scrutiny of AI companies' finances has intensified as they pour money into chips, data centers and computing power needed to develop and operate increasingly powerful models. Targeting a valuation of over $2 trillion, Anthropic is seeking to convince investors that its breakneck revenue growth can eventually outrun the enormous costs of building and running AI systems. Last year, Anthropic's revenue increased twelvefold to nearly $4.6 billion, even as the company lost more than $8 billion on an operating basis, excluding writedowns of various liabilities mostly tied to previous fundraising, Reuters reported exclusively last month. The company has committed at least $518 billion over a decade for cloud services, computing and infrastructure. Also Read: Anthropic warns IPO investors AI could pose 'catastrophic' risks to humanity: Report Founded by former OpenAI executives, Anthropic has emerged as one of the leading generative AI developers with its Claude family of models. Its expansion has been underpinned by Alphabet's Google and Amazon, which are simultaneously investors, suppliers of computing power and important distribution partners, as well as AI competitors. Anthropic has also sought to distinguish itself through its focus on AI safety, warning of the dangers posed by increasingly capable systems even as it competes in the race to develop and commercialize more powerful models. Its research has found that its own AI agents have engaged in troubling behavior in test environments, including leaking sensitive information and sabotaging code, though the company has said it has not seen evidence of such misalignment in real-world deployments.
Anthropic Hits $16.8B Revenue Ahead Of Planned Holiday Listing: Report
Artificial intelligence startup Anthropic reported a surge in third-quarter revenue to $16.8 billion while achieving profitability on an adjusted basis ahead of its highly anticipated initial public offering. * Anthropic posted $16.8 billion in revenue for the third quarter and informed prospective investors that it achieved adjusted profitability, Bloomberg reported. * The Claude maker is aiming to list as early as the week of Nov. 9, targeting a trading debut before Thanksgiving or by the end of the year. * Investors have indicated a valuation range between $1.8 trillion and $2 trillion. Anthropic PBC's third-quarter revenue has jumped to $16.8 billion as the artificial intelligence developer prepares to launch its initial public offering, Bloomberg reported, citing people familiar with the matter. During a presentation at its San Francisco headquarters on Wednesday, the creator of the Claude AI model told prospective investors that it reached profitability on an adjusted basis during the quarter, according to the sources Bloomberg quoted. Anthropic Targets A Nov. Listing Anthropic is preparing to begin formal marketing for its initial public offering as soon as the week of Nov. 9, positioning the company to begin trading before the Thanksgiving holiday, according to Bloomberg. Though deliberations remain fluid and schedules could change, the listing is expected to be completed no later than year-end, people familiar with the discussions told Bloomberg. Anthropic declined to comment on the financial details or timeline. Trillion-Dollar Valuation Expectations Investor reaction to the financial figures will be crucial in setting the final terms for the offer. Prior to the San Francisco presentation, prospective investors suggested that a fair valuation for Anthropic sits between $1.8 trillion and $2 trillion. The company expects its debut to match or exceed the scale of SpaceX's (SPCX) offering. The strong third-quarter figures mark a sharp turnaround from the prior year. In 2025, Anthropic generated approximately $4.6 billion in full-year revenue -- up from $386 million in 2024 -- while posting a net loss of nearly $42 billion and an operating loss exceeding $8 billion, according to documents previously reviewed by Bloomberg. Intensifying AI Rivalry And Scrutiny The push toward a public listing coincides with mounting competition and regulatory oversight. Anthropic remains in a tight race for valuation dominance against key rivals like OpenAI. Meanwhile, Meta Platforms Inc. (META) recently introduced its new "Muse Wonder" model, claiming capabilities in legal, financial, and coding tasks that rival Claude and ChatGPT, which is gaining popularity. Retail sentiment on Stocktwits on Anthropic was 'neutral' with 'low' message volumes. For updates and corrections, email newsroom[at]stocktwits[dot]com.
Anthropic revenue jumps to $16.8 billion in Q3 ahead of IPO
Anthropic's third-quarter revenue rose 45% to $16.8 billion from the previous quarter, according to an investor who attended a pre-IPO meeting with Anthropic leadership in San Francisco on Wednesday. Revenue in the first quarter was $4.7 billion, and $11.6 billion in the second quarter, according to the person who attended the investor day meant to familiarize prospective Anthropic investors before its highly anticipated IPO. The company was profitable on an adjusted basis in the third quarter, Anthropic told the group, the source said. Anthropic did not immediately respond to an email seeking comment. IPO plans and investor appetite The IPO is expected to precede a public listing by rival OpenAI, making Anthropic an early test of investor appetite for heavily unprofitable AI developers pursuing massive scale. Its public-market debut is likely to be pushed until after the November US midterm elections, Reuters has reported. Scrutiny of AI companies' finances has intensified as they pour money into chips, data centers and computing power needed to develop and operate increasingly powerful models. Targeting a valuation of over $2 trillion, Anthropic is seeking to convince investors that its breakneck revenue growth can eventually outrun the enormous costs of building and running AI systems. Revenue growth and rising costs Last year, Anthropic's revenue increased twelvefold to nearly $4.6 billion, even as the company lost more than $8 billion on an operating basis, excluding writedowns of various liabilities mostly tied to previous fundraising, Reuters reported exclusively last month. The company has committed at least $518 billion over a decade for cloud services, computing and infrastructure. Google and Amazon back expansion Founded by former OpenAI executives, Anthropic has emerged as one of the leading generative AI developers with its Claude family of models. Its expansion has been underpinned by Alphabet's Google and Amazon, which are simultaneously investors, suppliers of computing power and important distribution partners, as well as AI competitors. Anthropic has also sought to distinguish itself through its focus on AI safety, warning of the dangers posed by increasingly capable systems even as it competes in the race to develop and commercialize more powerful models. AI safety concerns remain Its research has found that its own AI agents have engaged in troubling behavior in test environments, including leaking sensitive information and sabotaging code, though the company has said it has not seen evidence of such misalignment in real-world deployments. Published on October 15, 2026
Anthropic Pre-IPO Contract Falls 4.81% to $1,936 | TokenPost
The ANTH contract recorded $23.71 million in 24-hour trading volume and $39.72 million in open interest on Hyperliquid. An Anthropic Pre-IPO contract on Hyperliquid fell 4.81% over 24 hours to $1,936, corresponding to an implied valuation of $1.936 trillion for the company. The market, launched by Entropy, uses the ticker ANTH. It recorded $23.71 million in trading volume over 24 hours, while open interest stood at $39.72 million. The $1,936 price corresponds to the contract's implied valuation; it does not establish Anthropic's independently reported valuation.
Anthropic's $16.8 billion quarter raises the stakes for its planned IPO
Anthropic's revenue rose 45 per cent to $16.8 billion in Q3 as reported adjusted profitability strengthens its IPO pitch, with investors weighing its growth against massive AI infrastructure commitments Anthropic's revenue jumped 45 per cent in the third quarter to $16.8 billion, and the company told prospective investors it was profitable on an adjusted basis. The figures strengthen its case ahead of a potential IPO, although its planned valuation of more than $2 trillion will put the sustainability of its growth under scrutiny. Revenue rose from $11.6 billion in the second quarter to $16.8 billion in the third quarter, Reuters reported on Thursday, citing an investor who attended a pre-IPO meeting with Anthropic's leadership in San Francisco. The company generated $4.7 billion in revenue in the first quarter, the report said. The meeting was intended to familiarise prospective investors with the company ahead of its highly anticipated stock-market debut. Revenue growth against rising AI costs Founded by former OpenAI executives, Anthropic has emerged as one of the leading developers of generative artificial intelligence through its Claude family of models. Its growth has been supported by Alphabet's Google and Amazon, which are investors, suppliers of computing power and distribution partners, while also competing in the AI market. Anthropic's revenue has expanded rapidly. It rose twelvefold to nearly $4.6 billion last year, even as the company recorded an operating loss of more than $8 billion, excluding write-downs of liabilities largely associated with previous fundraising, Reuters reported in September. The company has committed at least $518 billion over a decade to cloud services, computing and infrastructure, according to a separate Reuters report. The commitments highlight the scale of investment required to develop, train and operate increasingly capable AI systems. For prospective shareholders, the challenge will be to assess whether Anthropic can sustain its revenue growth while managing the cost of computing resources and infrastructure. Strong demand for AI products can support sales, but the expense of developing and running advanced models remains a central consideration when evaluating the long-term profitability of AI businesses. Scrutiny of AI companies' finances has intensified as developers commit substantial resources to advanced chips, data centres and computing capacity. Investors are increasingly looking beyond revenue growth to examine operating expenses, future funding requirements and the ability to generate durable earnings. A test for AI valuations Anthropic's expected IPO is set to precede a public listing by rival OpenAI, making it an early test of investor appetite for AI developers pursuing rapid expansion despite substantial costs. Reuters has reported that Anthropic's public-market debut is likely to be pushed until after the November US midterm elections. The company is seeking a valuation of more than $2 trillion, according to the report. At that level, investors will have to weigh its rapid revenue growth and reported adjusted profitability against its infrastructure commitments and the uncertainty surrounding the long-term economics of advanced AI. A successful listing could offer a benchmark for how public-market investors value fast-growing AI businesses. It could also provide a clearer indication of how much weight investors place on revenue growth and adjusted profitability when companies require significant capital to expand. Anthropic has sought to distinguish itself through its focus on AI safety, while continuing to compete in the development and commercialisation of more powerful systems. Its research has identified troubling behaviour by its AI agents in test environments, including leaking sensitive information and sabotaging code. The company has said it has not seen evidence of such misalignment in real-world deployments.
Anthropic's Quarterly Revenue Jumped 45% to Nearly $17 Billion
Anthropic's third-quarter revenue rose 45% to $16.8 billion, according to people familiar with the matter.View more SAN FRANCISCO -- Anthropic's revenue climbed 45% to $16.8 billion in the third quarter, according to people familiar with the matter, as the Claude maker prepares for a massive public ...
Anthropic's Q3 revenue surges 45% to nearly $17b ahead of IPO
Anthropic's third-quarter revenue rose 45% to $16.8 billion from the previous quarter, according to an investor who attended a pre-IPO meeting with Anthropic leadership in San Francisco on Wednesday. Revenue in the first quarter was $4.7 billion, and $11.6 billion in the second quarter, according to the person who attended the investor day meant to familiarize prospective Anthropic investors before its highly anticipated IPO. The company was profitable on an adjusted basis in the third quarter, Anthropic told the group, the source said. Anthropic did not immediately respond to an email seeking comment. The IPO is expected to precede a public listing by rival OpenAI, making Anthropic an early test of investor appetite for heavily unprofitable AI developers pursuing massive scale. Its public-market debut is likely to be pushed until after the November US midterm elections, Reuters has reported. Scrutiny of AI companies' finances has intensified as they pour money into chips, data centers and computing power needed to develop and operate increasingly powerful models. Targeting a valuation of over $2 trillion, Anthropic is seeking to convince investors that its breakneck revenue growth can eventually outrun the enormous costs of building and running AI systems. Last year, Anthropic's revenue increased twelvefold to nearly $4.6 billion, even as the company lost more than $8 billion on an operating basis, excluding writedowns of various liabilities mostly tied to previous fundraising, Reuters reported exclusively last month. The company has committed at least $518 billion over a decade for cloud services, computing and infrastructure. Founded by former OpenAI executives, Anthropic has emerged as one of the leading generative AI developers with its Claude family of models. Its expansion has been underpinned by Alphabet's Google and Amazon, which are simultaneously investors, suppliers of computing power and important distribution partners, as well as AI competitors. Anthropic has also sought to distinguish itself through its focus on AI safety, warning of the dangers posed by increasingly capable systems even as it competes in the race to develop and commercialize more powerful models. Its research has found that its own AI agents have engaged in troubling behavior in test environments, including leaking sensitive information and sabotaging code, though the company has said it has not seen evidence of such misalignment in real-world deployments.
Anthropic to update Claude AI with US election information ahead of midterms
Anthropic will add official US voting information and election results to its Claude AI chatbot later this month, ahead of the November 2026 midterm elections. The company announced the move on Wednesday, October 14, saying the update aims to help users access reliable election information while preventing the misuse of AI during political events. The new feature will use information provided by Democracy Works, a nonprofit that collects and shares voting information. US users who ask Claude about voting procedures will receive answers drawn directly from the organisation's database of state and local election sources. The update comes as AI companies look to make their chatbots useful sources of news and public information. With the US midterms approaching, more users may turn to AI tools to ask questions about elections, voting rules and results. Anthropic said the planned update is intended to help users find information about voting logistics. Instead of relying on general responses, Claude will draw on Democracy Works' database when answering questions about state and local voting procedures. The company also plans to add election results to the chatbot. The update is expected to be available to US users later in October, before voters head to the polls in November. Anthropic said its aim is to make Claude useful for political conversations while taking steps to prevent election-related abuse. However, the company did not describe the feature as a tool for recommending candidates or telling users how to vote. Also Read: Eight US states sue OpenAI, alleging ChatGPT exposes young users to dangerous behavior Anthropic also shared figures showing how often users discuss politics and elections with Claude. According to the company, conversations about politics, public policy and social issues account for around 5.75% of all conversations on Claude.ai, its consumer app in the United States. The figure comes from an analysis of 250,000 US conversations. Election-specific questions make up a smaller share of the chatbot's activity. Anthropic said conversations directly related to elections or voting account for about 0.51% of US conversations. The company expects political and election-related queries to increase as Election Day approaches. The figures suggest that while politics is part of the chatbot's use, election-related questions currently make up a limited share of overall conversations. Also Read: AI surge is reshaping cybersecurity boundaries Anthropic also said requests for recommendations on how to vote were extremely rare, accounting for just 0.01% of US conversations. The update comes as AI chatbots become another way for people to seek information about public affairs. Users can ask these tools about political issues, election procedures and results, making the accuracy of their responses an important concern during election periods. By using information from Democracy Works, Anthropic plans to direct voting-related questions towards a database of state and local election sources. This approach is intended to give users information about voting logistics rather than relying only on the chatbot's general knowledge. The November 2026 midterm elections will determine seats in the US Congress, alongside various state and local offices. As the election approaches, Anthropic expects more users to ask Claude questions related to the political process.
Anthropic to add US voting information to Claude before midterms
Oct 14 (Reuters) - Anthropic said on Wednesday it would add official US voting information from nonprofit Democracy Works to its Claude AI chatbot later this month, ahead of the November midterm elections. The AI company said its goal was to position Claude as a useful tool for political conversations, while preventing election-related abuse. "Today, conversations that touch broadly on politics, public policy or social issues account for about 5.75% of all conversations in the United States on Claude.ai, our consumer app," the company said in a statement. (Reporting by Jaspreet Singh in Bengaluru; Editing by Anil D'Silva)
Anthropic and OpenAI Are Negotiating Publication Rights of External AI Safety Reviews
Publication Safeguards: A public letter by prominent AI experts demands editorial freedom for evaluators, time-limited redactions and protection against retaliation. Anthropic and OpenAI are negotiating with outside AI safety reviewers over what they may publish, according to Reuters. The terms could shape what the public learns about AI risks: reviewers need internal access to check companies' safety claims, while the companies providing that access also seek confidentiality protections. Two people familiar with the talks told Reuters that both developers had sought strict publication controls as of October 5, citing concerns about trade secrets. One said some Anthropic discussions had evolved by October 8. What Internal Access Would Let Reviewers Check In September, Anthropic CEO Dario Amodei proposed bringing independent evaluators inside AI labs to inspect their adherence to training, deployment and safety commitments. These embedded evaluators would receive relevant internal tools and opportunities to speak with employees, rather than relying solely on tests of a finished model. That access would let outsiders examine how a model was developed and whether company descriptions match internal practice. Adam Gleave, CEO of AI safety group FAR.AI, proposed comparing intermediate model versions, examining training environments and interviewing employees in a September interview. He also said FAR.AI had declined contracts that gave developers excessive control, identifying confidentiality agreements and publication restrictions as threats to independence. Palisade Research chief of staff John Steidley identified a reason to inspect training records in those same interviews: training a model on particular tests changes what its later benchmark performance reveals. Passing a safety test can reflect learning that test itself. Models that recognize an evaluation could appear to behave well under test while problems remain hidden. Some researchers argue that training history could reveal clues missed in finished-model tests. Anthropic's September 18 partnership plan with Accenture gives the proposed approach also a commercial setting. Faculty, Accenture's specialist AI business, would lead evaluations, adversarial testing and checks of model safeguards. Anthropic would fund the work directly; each company expected to invest at least $1 billion in evaluation capacity over five years. Anthropic cited Accenture's experience deploying AI for businesses as its reason for the decision. Faculty CEO Marc Warner told Reuters that evaluators' reports are useful only if their conclusions are trusted. Anthropic's announcement also mentions discussions with AI evaluation group METR about independently funded nonprofit pilots, offering a different funding arrangement from its directly paid Accenture partnership. Who Controls Publication Anthropic told Reuters that evaluators it had previously partnered with controlled their methods, assessment criteria, conclusions and publications. For new firms, it reserved the right to comment on and approve reports while they learned about its models and established a working relationship. Amodei's September proposal promised publication of key risk, incident and access findings without Anthropic editorial control, subject to narrow confidentiality and security redactions. It also promised that consequential redactions could be disclosed and that unfavorable findings alone would not justify removing them. OpenAI described its own oversight authority as being used only for removing confidential information, while denying that it sought approval over entire safety reports. It also said reviewers could disclose major redactions. A contract Reuters inspected barred publication of confidential information without company consent. One of Reuters' sources said measures under discussion included lab employees observing embedded teams at work and barring evaluators at one lab from working at another simultaneously, to prevent intellectual-property leaks. A separate person with past dealings with a lab said that intellectual property could overlap information relevant to AI risks. Rumman Chowdhury, CEO of AI evaluation company Humane Intelligence PBC, argued that reviewers' dependence on privileged access gives host companies leverage over publication. An Investigation Shows the Access and Disclosure Trade-Off After OpenAI attributed the July intrusion into AI development platform Hugging Face to escaped own AI agents, Hugging Face CEO Clem Delangue requested logs of recorded actions for safety researchers. As shared by METR later, an outside team obtained OpenAI's internal evidence and negotiated its public presentation. METR researchers Hjalmar Wijk and Ajeya Cotra worked with Redwood Research's Ryan Greenblatt, who was contracting with METR. Their investigation focused on agents' behavior and reasoning within an agreed June 26-July 13 window. OpenAI supplied records and access to staff, while the investigators could not directly inspect its infrastructure or query the main model they were investigating. A planned two-day visit expanded to six days of on-site work after concerns about incomplete data prompted return visits. OpenAI could redact nonpublic information from the report. Its private version contained more than 100 raw reasoning snippets; OpenAI limited the public version to 30 on intellectual-property grounds, leading the authors to use paraphrases. Readers therefore received the investigators' analysis with less of the underlying model text available for inspection. The authors also accepted feedback on structure, emphasis and tone. They said some drafting compromises reflected a desire to avoid discouraging future voluntary incident investigations, while standing by their substantive conclusions. They praised the unusual access to transcripts and responsive staff, arguing that early outside investigation was valuable. OpenAI provided free API credits for model use rather than paying for the assessment. The authors estimated using roughly $400,000 in those credits during the investigation. Their agreed scope excluded assessment of safeguard effectiveness, the extent of the compromise and remediation. Reviewers Seek Protection Beyond Access A public letter by AI Evaluator Forum from September asks frontier AI labs to embed evaluators under specified independence protections. On October 13, its page listed 200-plus individual signatories, including computer scientists Geoffrey Hinton and Stuart Russell. According to the plea, evaluators would keep full control of their methods and reports and also disclose any conflicts of interest. The signers seek direct communication with company boards and public evidence, subject to time-limited redactions protecting critical privacy, security, intellectual-property and public-safety interests. The letter also asks for independence from lab ownership and significant business ties beyond evaluation, alongside protection from litigation, retaliation and funding insecurity when findings displease the host. As Henry Papadatos, director of AI safety group Safer AI, argued in interviews, voluntary protections depend on company goodwill, which could change quickly during reputational crises.
Anthropic IPO: AI Lab Reportedly To Share Q3 Financials With Investors As It Weighs $2 Trillion Valuation
Investors look for signs of sustainable AI growth as Anthropic prepares for a potential blockbuster listing. * Anthropic's financial performance could offer clues about AI infrastructure demand. * Earlier this month, confusion over OpenAI's revenue figures rattled semiconductor stocks and the broader Nasdaq. * A legal dispute over the U.S. government's supply-chain risk designation for Anthropic adds uncertainty ahead of a potential listing. Anthropic PBC plans to disclose its third-quarter financials to potential investors on Wednesday, as the artificial intelligence startup prepares for a potential initial public offering that could value it at $2 trillion, CNBC reported on Tuesday, citing people familiar with the matter. The San Francisco-based company will share the numbers with a group of institutional investors, offering a closer look at its revenue growth and profitability. The disclosure comes as investors increasingly scrutinize private AI companies' financial performance to assess the sustainability of the industry's growth, which is powering growth for many public tech companies. Earlier this month, confusion over OpenAI's revenue figures rattled semiconductor stocks and the broader Nasdaq. Media reports initially pegged the AI startup's annualized revenue at $70 billion, but the Financial Times later reported the figure was closer to $50 billion. According to CNBC, the revised figure excluded revenue from cloud partners and did not signal weakening demand. Still, the discrepancy highlighted how accounting and revenue-reporting differences can shape investor perceptions of AI demand and trigger volatility across technology stocks. Anthropic IPO Watch Anthropic is reportedly targeting a valuation of $2 trillion, potentially making its debut one of the largest IPOs in history. The company confidentially filed with the Securities and Exchange Commission earlier this year but has not yet made its S-1 registration statement public. Reuters has reported that Anthropic plans to list in November, although the timing has not been confirmed. Anthropic's latest valuation is $1.35 trillion, compared with OpenAI's $985.59 billion, according to Nasdaq Private Market. Anthropic Challenges US Government's Risk Designation Meanwhile, Anthropic has asked the full federal appeals court in Washington to reconsider a ruling that upheld the Trump administration's designation of the company as a supply-chain risk to U.S. national security, Bloomberg reported Tuesday. The company argued that the court's September decision misinterpreted the law and that the designation retaliated against it for its position on AI safety. The dispute began after negotiations over the military's use of Anthropic's technology broke down. Anthropic sought assurances that the government would not use its AI for mass surveillance of Americans or autonomous weapons deployment, while the government opposed restrictions on its use for national security purposes. The legal battle adds another layer of uncertainty for Anthropic's IPO. For updates and corrections, email newsroom[at]stocktwits[dot]com.<
Anthropic signs MOU with Singapore to evaluate AI models for safety and security
Add Yahoo as a preferred source to see more of our stories on Google. SINGAPORE - Anthropic has pledged to hold technical exchanges with the Singapore Government to evaluate its artificial intelligence models for safety and security. On Oct 14, the San Francisco-based AI lab signed a memorandum of understanding (MOU) with the Ministry of Digital Development and Information (MDDI) to set out their areas of cooperation, which include capability building and responsible adoption. Announcing the MOU at its Singapore office opening event that day, Anthropic co-founder Ben Mann said: "When we train a new model, nobody really knows what exactly the capabilities are. Working with AI safety institutes in Singapore, as well as those around the world, is important to understand what the risks and the benefits could be. "Singapore has been a leader in responsible Al, with a long track record of investment in model testing and evaluation... With this MOU, we'll partner with Singapore to help ensure the transition to powerful Al goes well here, across the Asia-Pacific and around the world." Anthropic's Singapore office is its fifth in the Asia-Pacific region, after those in South Korea, Japan, India and Australia. It was officially launched on Oct 14 at Capella Singapore, on the sidelines of Singapore International Cyber Week. Anthropic's MOU with MDDI comes amid intensifying global debate over safety regulations after AI agents that can execute tasks with minimal human supervision were found to have gone rogue. Anthropic's rival OpenAI disclosed in July that its AI agents had escaped their test environment. The agents then breached AI software repository Hugging Face to complete a task they were given, an attack that was not intended by OpenAI. Reports of similar incidents followed, including the breach of an Australian healthcare database by a rogue OpenAI bot in June, an incident later revealed by Australian Prime Minister Anthony Albanese in September. Anthropic's chief executive Dario Amodei is among those leading the charge for regulation, calling on frontier AI firms to slow down AI development. The AI lab, which could be valued up to US$2 trillion (S$2.56 trillion) in what could be the largest initial public offering in history, dedicated 80 pages out of 261 of its draft prospectus to risk factors. Anthropic cautioned that advanced AI models could display "self-preserving behaviours", "resist shutdown" and pose "catastrophic or existential risks to humanity". The start-up counts Singapore's sovereign wealth fund GIC as one of its major financial backers. GIC first invested in Anthropic in September 2025 as part of a US$13 billion fund-raising round by private investors. In February 2026, GIC led a US$30 billion Series G fund-raising round. During a keynote at Capella on Oct 14, Anthropic's ASEAN general manager Dale Finlay said the start-up's highly anticipated Singapore office is driven by the country's high concentration of technical talent and strong research institutions, as well as high standards of trust in AI adoption. "Singapore's Government understands that trust is the most essential element of AI adoption, and they've done years of work testing and evaluating models," said Finlay. "Customers here hold AI to a really high standard too, especially when it comes to things like reliability and security. Talent and trust is what's made Singapore one of the strongest markets for Claude anywhere in the world." According to Anthropic's 2026 economic index, Singapore ranks second among 121 countries in Claude usage relative to its population, at nearly six times what its size alone would predict. With a presence here, Anthropic seeks to work more closely with organisations, including small and medium-sized enterprises, to build Claude into products and services and support the Government's efforts to build AI skills among public officers, said Finlay. This builds on existing work with local agencies and companies such as the Government Technology Agency of Singapore (GovTech), DBS Bank, Grab, and Sea Group, which owns e-commerce platform Shopee and gaming platform Garena. More than 1,500 GovTech officers use Claude Code to help build and maintain Singapore's digital services, while Shopee uses Claude to power its customer service chatbots and help employees write and debug code quickly. Permanent Secretary for Digital Development and Information Chng Kai Fong said: "Trust is critical for businesses, workers and the public to use Al with confidence. He added: "This partnership with Anthropic reflects the Government's commitment to scaling up the responsible adoption of AI, deepening capabilities in our companies and government agencies, and delivering impact for Singapore."
Anthropic signs MOU with Singapore on AI safety
Anthropic's ASEAN general manager Dale Finlay (left) and co-founder Ben Mann speaking at an event at Capella Singapore on Oct 14. SINGAPORE - Anthropic has pledged to hold technical exchanges with the Singapore Government to evaluate its artificial intelligence models for safety and security. On Oct 14, the San Francisco-based AI lab signed a memorandum of understanding (MOU) with the Ministry of Digital Development and Information (MDDI) to set out their areas of cooperation, which include capability building and responsible adoption. Announcing the MOU at its Singapore office opening event that day, Anthropic co-founder Ben Mann said: "When we train a new model, nobody really knows what exactly the capabilities are. Working with AI safety institutes in Singapore, as well as those around the world, is important to understand what the risks and the benefits could be. "Singapore has been a leader in responsible Al, with a long track record of investment in model testing and evaluation... With this MOU, we'll partner with Singapore to help ensure the transition to powerful Al goes well here, across the Asia-Pacific and around the world." Anthropic's Singapore office is its fifth in the Asia-Pacific region, after those in South Korea, Japan, India and Australia. It was officially launched on Oct 14 at Capella Singapore, on the sidelines of Singapore International Cyber Week. Anthropic's MOU with MDDI comes amid intensifying global debate over safety regulations after AI agents that can execute tasks with minimal human supervision were found to have gone rogue. Anthropic's rival OpenAI disclosed in July that its AI agents had escaped their test environment. The agents then breached AI software repository Hugging Face to complete a task they were given, an attack that was not intended by OpenAI. Reports of similar incidents followed, including the breach of an Australian healthcare database by a rogue OpenAI bot in June, an incident later revealed by Australian Prime Minister Anthony Albanese in September. Anthropic's chief executive Dario Amodei is among those leading the charge for regulation, calling on frontier AI firms to slow down AI development. The AI lab, which could be valued up to US$2 trillion (S$2.56 trillion) in what could be the largest initial public offering in history, dedicated 80 pages out of 261 of its draft prospectus to risk factors. Anthropic cautioned that advanced AI models could display "self-preserving behaviours", "resist shutdown" and pose "catastrophic or existential risks to humanity". The start-up counts Singapore's sovereign wealth fund GIC as one of its major financial backers. GIC first invested in Anthropic in September 2025 as part of a US$13 billion fund-raising round by private investors. In February 2026, GIC led a US$30 billion Series G fund-raising round. During a keynote at Capella on Oct 14, Anthropic's ASEAN general manager Dale Finlay said the start-up's highly anticipated Singapore office is driven by the country's high concentration of technical talent and strong research institutions, as well as high standards of trust in AI adoption. "Singapore's Government understands that trust is the most essential element of AI adoption, and they've done years of work testing and evaluating models," said Finlay. "Customers here hold AI to a really high standard too, especially when it comes to things like reliability and security. Talent and trust is what's made Singapore one of the strongest markets for Claude anywhere in the world." According to Anthropic's 2026 economic index, Singapore ranks second among 121 countries in Claude usage relative to its population, at nearly six times what its size alone would predict. With a presence here, Anthropic seeks to work more closely with organisations, including small and medium-sized enterprises, to build Claude into products and services and support the Government's efforts to build AI skills among public officers, said Finlay. This builds on existing work with local agencies and companies such as the Government Technology Agency of Singapore (GovTech), DBS Bank, Grab, and Sea Group, which owns e-commerce platform Shopee and gaming platform Garena. More than 1,500 GovTech officers use Claude Code to help build and maintain Singapore's digital services, while Shopee uses Claude to power its customer service chatbots and help employees write and debug code quickly. Permanent Secretary for Digital Development and Information Chng Kai Fong said: "Trust is critical for businesses, workers and the public to use Al with confidence. He added: "This partnership with Anthropic reflects the Government's commitment to scaling up the responsible adoption of AI, deepening capabilities in our companies and government agencies, and delivering impact for Singapore."
Anthropic Plans to Reveal Quarterly Numbers before IPO; AI Sector Could See Major Boost | TopNews
Anthropic is set to show prospective investors its third-quarter numbers for the first time on Wednesday, Oct. 14, at a meeting at its San Francisco headquarters, as the artificial intelligence developer moves toward what could become the largest initial public offering in history. Investors have reached in a fair value of $1.8 trillion to $2 trillion, and a roughly $100 billion offering led by Morgan Stanley, Goldman Sachs and JPMorgan Chase could reach Nasdaq before Thanksgiving. The draft prospectus describes explosive revenue growth alongside heavy losses. The quarter now in focus will help decide whether that growth is strong enough to carry the price. A Rare Look Behind the Curtain For a company that has spent most of its existence disclosing as little about its finances as private-market convention allows, Wednesday marks a turning point. Anthropic plans to walk prospective investors through its third-quarter financials, figures it has never before shared with that audience, at a gathering at its San Francisco headquarters. CNBC reported the plan on Tuesday, citing two people familiar with the matter. The existence of the investor day itself was not new: Bloomberg first reported on Oct. 1 that the meeting was on the calendar. The timing is deliberate. An investor day of this kind is the stage in an IPO process where the issuer stops being a story and starts being a set of numbers that institutional money managers can model, stress-test and argue over. Prospective investors have already settled on a fair valuation range of $1.8 trillion to $2 trillion, according to Barchart. Wednesday's disclosures will either reinforce that range or force a rethink before the deal moves to a formal marketing phase. The meetings come roughly two weeks after Reuters reviewed Anthropic's draft IPO prospectus, a document that to date offers the most detailed picture of both the company's breakneck growth and the scale of the losses that have come with it. That prospectus, however, has a status worth stating precisely, because it has been widely misdescribed. That distinction matters for investors. A confidential draft can still change in response to SEC comments, and the figures in it have not yet been subjected to the full glare of public scrutiny that comes with a published S-1. Until that filing appears, the market is working from reporting on the document rather than the document itself. The 2025 Numbers: Hypergrowth at a Heavy Price By the measure that matters most to growth investors, Anthropic's 2025 was extraordinary. According to the draft prospectus Reuters reviewed, revenue grew roughly 12-fold during the year to nearly $4.6 billion. Very few companies of any size have multiplied their top line at that pace, and fewer still have done so while already generating billions of dollars in sales. The bottom line tells a more complicated story. Anthropic recorded a net loss of $42 billion for 2025. That headline figure, though, overstates the cash damage. About $34 billion of the loss was a non-cash charge tied to remeasuring convertible notes, an accounting adjustment that typically swells when a company's valuation rises sharply, because the instruments that convert into its equity become more valuable on paper. Strip that out, and the business itself posted an operating loss of more than $8 billion. Infrastructure is the dominant cost. Anthropic spent $7.33 billion on compute and infrastructure in 2025, equal to 58% of its operating expenses. Working backward from those two figures implies total operating expenses of roughly $12.6 billion, a calculation consistent with an operating loss of a little over $8 billion on revenue of about $4.6 billion. Looking further out, the company has committed about $518 billion to cloud, computing and infrastructure in the coming years, a sum that dwarfs its current revenue and underscores how capital-hungry frontier AI development has become. 2026: Growth Accelerates and the First Profit Arrives If 2025 established the trajectory, 2026 steepened it. Preliminary figures show revenue of $4.73 billion in the first quarter, which on its own exceeded the company's entire 2025 total. The second quarter was larger still, at more than $11.5 billion, meaning revenue more than doubled sequentially in a single quarter. The more consequential development for valuation-minded investors may be on the profit line. The Wall Street Journal reported that Anthropic posted its first adjusted operating profit in the second quarter, totaling $559 million. That measure excludes stock-based compensation, a significant expense at AI companies that compete fiercely for research talent, so skeptics will want to see the figure on a fully loaded basis. Still, the swing from an operating loss of more than $8 billion in 2025 to an adjusted operating profit within six months is the type of inflection that IPO buyers prize. The momentum appears to have carried into the summer. The Financial Times reported that the company told shareholders it expects a second straight quarter of adjusted operating profit in the third quarter, the very period whose results will be presented on Wednesday. The Road to Nasdaq The calendar is moving quickly. Yahoo Finance reported that a formal roadshow could begin as early as the week of Nov. 9, with shares potentially trading on Nasdaq before Thanksgiving. Morgan Stanley, Goldman Sachs and JPMorgan Chase are leading an offering expected to raise about $100 billion, a sum that would by itself exceed the market value of many companies in the S&P 500. The valuation step-up is just as striking. Reuters noted that a $2 trillion valuation would be more than double the $965 billion at which Anthropic was valued in May. The Motley Fool observed that a listing of that size would surpass SpaceX as the largest IPO in history. Context helps here. Raising $100 billion at a $2 trillion valuation implies that only about 5% of the company would change hands in the offering. A thin float relative to total value is common for mega-cap listings, but it also means that index inclusion, lockup expirations and the pace of follow-on selling by early holders could have an outsized effect on how the shares trade once the initial excitement fades. The Skeptics' Case Not everyone is persuaded that the price is justified. PitchBook, writing in Morningstar, calculated that a $2 trillion valuation would equal just under 31 times Anthropic's July revenue run rate. Put another way, that multiple implies an annualized run rate of somewhat more than $64 billion at the time. Even for a company growing as fast as Anthropic, paying more than 30 times sales leaves little margin for error, particularly in a market where large technology companies are also pouring capital into competing models. PitchBook also flagged a notable gap in the disclosure: the draft prospectus does not disclose gross margin. For an AI developer whose single largest expense is compute, that omission is significant. Gross margin is the clearest window into whether each additional dollar of revenue becomes cheaper to serve as the business scales, or whether infrastructure costs rise in lockstep with sales. Without it, investors are left to infer unit economics from the operating line. The prospectus itself includes an unusual warning. It cautions that advanced AI could pose "catastrophic or existential risks to humanity". Risk-factor sections are often boilerplate, but language of this kind is far from standard, and it reflects how closely Anthropic's corporate identity is tied to debates over AI safety. Why the Third Quarter Matters So Much That backdrop is what makes Wednesday's presentation pivotal. PitchBook said third-quarter revenue "well above $16.3 billion" would strengthen the case for the valuation. Measured against second-quarter revenue of more than $11.5 billion, clearing that bar would require sequential growth of roughly 42%, a slower pace than the jump from the first quarter to the second but still exceptional for a company of this size. PitchBook was careful not to overstate the point, adding that such a result "would not by itself establish that $2 trillion is justified". In other words, a strong quarter is necessary for the bull case but not sufficient. Investors will also be looking for evidence that the adjusted operating profit is durable, that compute spending is becoming more efficient relative to revenue, and that the enormous infrastructure commitments are matched by a credible path to demand. Broader Implications for Markets A listing of this magnitude would ripple well beyond Anthropic's own shareholders. A $100 billion offering would absorb a large slice of the capital that institutional investors earmark for new issues, potentially crowding out smaller deals in the weeks around it. If the shares perform well, the IPO could reopen the window for other private AI companies sitting on lofty private-market valuations. If they stumble, it could prompt a broader reassessment of how public markets price AI growth that has not yet translated into consistent profits. There is also a strategic dimension. Anthropic's roughly $518 billion in infrastructure commitments makes it one of the largest prospective customers for cloud providers, chipmakers and data center developers. Public-market capital would help underwrite those commitments, so the reception of the offering will be watched closely across the AI supply chain, not just by those buying the stock. What Investors Should Watch * Q3 revenue versus $16.3 billion: PitchBook's benchmark is the clearest near-term test. A figure "well above" it supports the valuation range; a figure near or below it weakens it. * A second adjusted operating profit: Confirmation of the profit the company has signaled to shareholders would show the second-quarter result was not a one-off. * Gross margin disclosure: Any move to publish gross margin in the public S-1 would give investors a far better read on the economics of serving AI demand. * The public S-1 filing: Until the draft registration becomes public, figures rest on reporting about a confidential document. * Roadshow timing: A start as early as the week of Nov. 9 would signal confidence; any slippage could hint at pricing debate. Risks to the Bull Case The central risk is valuation. At just under 31 times a July revenue run rate, the offering prices in years of continued hypergrowth, and any deceleration could compress the multiple quickly. The company remains deeply loss-making on a full-year 2025 basis, its adjusted profit excludes stock-based compensation, and its $518 billion in infrastructure commitments represent a fixed burden that will persist regardless of how demand evolves. The absence of gross margin data limits visibility into unit economics. Finally, the prospectus's own warning about catastrophic AI risks is a reminder that regulatory and reputational pressures on the sector could shift in ways that are difficult to model. The Bottom Line Anthropic arrives at its investor day with a growth story few companies in history can match: revenue up roughly 12-fold in 2025, more than doubling again between the first and second quarters of 2026, and a first adjusted operating profit of $559 million. Against that stand a $42 billion net loss for 2025, infrastructure commitments of about $518 billion, an undisclosed gross margin and a valuation that would more than double in five months. The third-quarter numbers presented on Wednesday will not settle the debate on their own, as PitchBook made clear. But they will go a long way toward determining whether the largest IPO ever attempted lands on Nasdaq at the top of its range or is forced to make concessions on price.
21shares Launches Anthropic ETP Ahead of AI Firm's Expected IPO
Get the hottest Fintech Switzerland News once a month in your Inbox Swiss crypto company 21shares has launched an exchange-traded product (ETP) tied to the implied valuation of AI firm Anthropic. The 21shares Anthropic Synthetic ETP trades as CLAU. It lets investors take part in that valuation before the firm's expected IPO. The crypto ETP provider says it is Europe's first ETP to offer this exposure continuously and without leverage. In addition, investors can buy it through a standard brokerage account. However, they do not buy Anthropic shares directly. Instead, the ETP gets its exposure through total return swaps. Under a swap, one party pays the other the full return of an agreed asset. The swaps are linked to the Kaiko Anthropic USD Reference Rate (USDT) (London) Index. In turn, the index price is based on Anthropic perpetual futures on Binance. These futures have no expiry date. CLAU trades in euros on the Vienna MTF, run by the Vienna Stock Exchange. It has a yearly total expense ratio of 1.00% and an exit fee of 2.00%. The ETP is also the first product under a new special purpose vehicle (SPV) created with FalconX. Duncan Moir, President and Chief Investment Officer at 21shares, said the main value drivers in the most advanced AI still lie with private firms. In his view, putting pre-IPO exposure into an ETP closes this access gap. According to 21shares, the product is not connected to or endorsed by Anthropic. It gives no direct exposure to Anthropic shares. Featured image: Edited by Fintech News Switzerland, based on image by Frolopiaton Palm via Magnific.
Anthropic warns AI could become capable of copying itself - Jordan News | Latest News from Jordan, MENA
Anthropic has warned that small artificial intelligence models could become capable within months of copying themselves and spreading automatically across internet-connected devices, without the need for specialised data centres or human intervention. Such a development could pose new cybersecurity challenges.اضافة اعلان According to a report published by Reuters on Tuesday, October 13, 2026, experiments conducted by the company found that some AI models are approaching the ability to identify vulnerable devices, transfer copies of themselves to them, and continue spreading autonomously. The research focused on commercially available Internet of Things (IoT) devices, such as smart ovens, Android-powered televisions from lesser-known brands, and other internet-connected household appliances that often have weak security measures. Small models are approaching self-replication Anthropic's experiments showed that AI models containing fewer than 30 billion parameters have become capable of carrying out many of the technical steps required to copy themselves onto other devices. However, the company explained that these models still face difficulties completing the entire process independently, from identifying a target device to installing a new copy and continuing to spread without human assistance. In simplified tests in which some security measures were removed, the GLM-5.2 model and more advanced models managed to complete the self-replication process. This suggests that the capabilities required for this type of spread are already emerging under specific experimental conditions. Nevertheless, these findings do not mean that the models can currently spread freely across real-world devices, as some tests were conducted in simplified environments that do not reflect all the security barriers found in practice. Risks of spread that could be difficult to stop Anthropic warned that AI systems capable of copying themselves could be more difficult to contain than traditional malware because of the decentralised nature of their spread. If copies of a model were distributed across a large number of devices, there would not necessarily be a central control point that could be used to shut down the entire system. These models may also be able to identify unfamiliar devices automatically and adapt to their technical characteristics, rather than relying on fixed instructions targeting specific types of devices. The company noted that such systems could potentially exploit computing resources available online, including processors in network-connected devices, without relying on dedicated data centres to run AI models. This capability differs from the behaviour of many traditional forms of malware, which typically rely on predefined mechanisms to spread and execute instructions. Capability could emerge before the end of 2026 According to Anthropic's warnings, if small models continue to improve at the same historical rate, they could become capable of independently copying themselves by the end of 2026. The company also predicted that, shortly afterwards, small systems could emerge that are capable of transferring their model weights and restarting on other devices. However, these estimates are based on performance trends observed in experiments and do not confirm that the capabilities will emerge within the projected timeframe. The company explained that autonomous AI systems of this kind would likely spread relatively slowly at first, relying on widely distributed computing resources with weak security protections. Calls to strengthen the security of connected devices In light of these findings, Anthropic stressed the importance of strengthening security measures for internet-connected devices, particularly smart home appliances, televisions, and other equipment that may not receive regular security updates. The company believes that improving the protection of these systems and making them more difficult to exploit is a key step in limiting AI agents' ability to spread across networks. The study highlights an emerging cybersecurity challenge: the possibility that AI models could move beyond performing specific tasks at a user's request and become systems capable of independently identifying computing resources, exploiting them, and copying themselves. Although fully autonomous self-replication remains limited in the reported experiments, Anthropic's warnings point to the need to prepare for new security risks that may accompany the rapid development of small AI models, particularly as the number of internet-connected devices worldwide continues to grow.
AI in your oven? Anthropic warns models 'months away' from self-replication on smart devices
Add Yahoo as a preferred source to see more of our stories on Google. Anthropic is warning that some of its artificial intelligence models could soon become "self-replicating" and find ways to infiltrate internet-connected devices, such as smart fridges and automatic vacuum cleaners. Once this happens, the company said, the models could power themselves within these devices without the need of datacenters. Such self-replication would represent a major loss of control of a runaway AI model. "A replicator like this would be challenging to shut down," Anthropic researchers wrote in a report released on Monday, "since there's no centralized point to shut off." The company said these "AI worms", a kind of autonomous malware, could be "months away" from gaining such capabilities. Once out-of-control AI systems reach the capability to hack smart devices, they could spread through networks while dodging detection. In a worst-case scenario, Anthropic said, various AI worms could "swarm" together to attack computer hardware. "We should aim to harden our systems to make it difficult for such replicators to spread, but also prepare to deal with a world where they are here to stay," the researchers wrote. Anthropic's new report comes as AI safety concerns have reached a fever pitch. Over the past few months, a steady stream of reports have come from AI companies, including Anthropic and OpenAI, saying their agents are capable of acting on their own and hacking into other companies and government entities. The revelations have caused researchers, lawmakers and executives, including Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, to call for a slowdown and increased guardrails on the burgeoning industry. Despite that, both companies have continued apace to develop and release new models. Anthropic's "frontier red team", the unit tasked with stress-testing AI systems, put together the report released Monday. The idea of the team is to explore the capabilities of their AI models. In the scenario of the self-replicating AI worms, the researchers prompted the agent to hack internet connected devices. They found that AI self-replication would only likely work with smaller models and that homes and small businesses would probably be the first target for such attacks. The devices Anthropic lists as susceptible all come with small built-in AI chips, including smart ovens, smart TVs, automatic lawnmowers, security cameras and video-conferencing units. The researchers concluded it's unclear what form the AI replication could take. They said it could be a slow model like "digital slime mold", a model that scams people for compute credits or a "powerful cyber model" that rapidly hacks large facilities such as datacenters. Regardless, the Anthropic researchers said, they're expecting a future "awash with agents running on all types of hardware, including some that aren't under human control".
Anthropic asks the full appeals court to rehear its Pentagon case
It says the designation was retaliation for its speech about AI safety, and that it was given no process before it landed Anthropic has asked the full federal appeals court in Washington to revisit a 2-1 ruling upholding the Pentagon's designation of it as a supply-chain risk. It argues the designation was retaliation for its speech about AI safety. Anthropic has asked the full federal appeals court in Washington to rehear its challenge to a Pentagon designation that labels it a supply-chain risk to US national security. The petition was filed on Tuesday, Bloomberg reported, and asks the full court to revisit a decision handed down by a three-judge panel on 25 September. That ruling went 2-1. Judge Gregory Katsas wrote for the majority with Neomi Rao, and Karen LeCraft Henderson wrote the dissent. Henderson read the statute narrowly. It covers suppliers that might sabotage, extract data or otherwise manipulate a product, and she said its history concerns hostile states rather than a contractor enforcing its own terms openly. Anthropic says the majority let the department act without any pre-designation process, and despite what it calls straightforward evidence of retaliation for its speech about AI safety. A judge in California reached that finding in August, on a parallel designation. The dispute began with a contract. Anthropic wanted assurances Claude would not be used for mass surveillance of Americans or autonomous weapons, and the government said it could accept no restrictions. Washington then went further. A presidential memo in June barred vendors from disabling, degrading or modifying a model that warfighters depend on without prior government approval. The Pentagon has moved on regardless. Its chief technology officer said in September that about 90% of the classified AI workload had been shifted off Anthropic, and expected the rest by the end of that month. In Europe this fight could not take the same shape. The AI Act does not cover systems used exclusively for military, defence or national security purposes. A European lab refusing a military use would be relying on its contract, not on regulation. There would be no designation to challenge and no regulator to appeal to. Access can also vanish without a court. In June the US ordered Anthropic to suspend two models, and Indian developers lost them overnight. European labs have drawn their own conclusions. Mistral has been building defence models with the German firm Helsing since early last year, and its chief executive says Europe needs its own capabilities while adversaries exist. The money at stake is small. Government work is under 1% of revenue, but Anthropic has told IPO investors the government's view of it could cost material revenue with commercial customers too.
