Anthropic Claude Mythos Cybersecurity Shock Rocks Crypto Titans

Anthropic Claude Mythos cybersecurity fears are rippling through the digital asset world as major crypto exchanges scramble for access to the company's most advanced AI model.

Anthropic's unreleased Claude Mythos Preview model is being tested only with a small group of partners under Project Glasswing, including large tech and security firms such as Amazon, Apple, Microsoft, Google and Nvidia.

In internal materials, Anthropic has warned that Mythos is "currently far ahead of any other AI model in cyber capabilities," and could exploit software vulnerabilities at a scale that defenders struggle to match.

The company says Mythos has already identified thousands of zero‑day vulnerabilities, some buried in widely used codebases for one to two decades.

According to reporting based on sources at The Information, major cryptocurrency exchanges including Coinbase and Binance have opened discussions with Anthropic in a bid to gain access to Mythos.

Both firms sit on billions of dollars in customer assets and have been frequent targets of hacking and data breaches, pushing them to adopt more aggressive cybersecurity testing.

Binance and custodian Fireblocks already use Anthropic's publicly available Claude Opus models for penetration testing, with Fireblocks reporting that the AI has spotted critical flaws missed by human testers.

Earlier evaluations of Opus 4.6 alone surfaced more than 500 previously unknown high‑severity bugs in open‑source libraries, underlining why crypto platforms see Mythos as both a powerful shield and a potential new weapon for attackers.

Rival OpenAI is also limiting access to a new cybersecurity‑focused model, offering it only to select partners through a controlled programme after warning of its advanced capability to find and exploit vulnerabilities.

Both Anthropic Claude Mythos cybersecurity controls and OpenAI's restricted rollout point to a new norm in frontier AI: powerful systems will be tested behind closed doors before wider deployment, as regulators and industry weigh the risks of unleashing tools that can break as much as they protect.

Crypto firms' push to tap Anthropic Claude Mythos cybersecurity power captures a broader tension: the same AI models that might finally lock down fragile financial infrastructure could also hand sophisticated attackers their most potent toolkit yet.