Australia partnered with Anthropic after regulators warned its Mythos AI found 271 Firefox bugs

In an effort to address potential cybersecurity vulnerabilities, the Australian government is partnering with Anthropic.

"Our government takes protection of critical infrastructure extremely seriously, which is why we're working with software providers and companies like Anthropic to make sure we are aware of emerging vulnerabilities," a spokesperson for Home Affairs Minister Tony Burke said in a statement on Thursday.

Earlier this week, Australian regulators sounded the alarm about Mythos, Anthropic's AI vulnerability detection tool.

The Australian Prudential Regulation ⁠Authority (APRA), the country's banking regulator, stated on Monday that it will "continue to assess the implications of technological advancements to ensure the ongoing safety and resilience of the financial system."

The Australian Securities and Investments Commission (ASIC) said it would closely monitor all developments regarding vulnerabilities that could potentially destabilize Australia's banking system.

Mythos is like a double-edged sword. It has the capability to detect vulnerabilities before hackers have the chance to exploit them. And it's very effective in doing so. For example, Mozilla found and fixed 271 vulnerabilities in the latest version of its web browser, Firefox, thanks to early access to Mythos.

On the other hand, experts have expressed concerns that Mythos can increase the pace of sophisticated cyberattacks and inflict significantly more damage to supply chains or vital infrastructure.

"This increases the risk that small, seemingly harmless bugs could, when combined, enable a serious attack. At the same time, there is a lack of public technical details to verify the full impact - it is plausible that real vulnerabilities are being exploited, but it is less clear how easily they can be exploited in practice," the National Cyber Security Center (NCSC) in the Netherlands said last week.

Like Australia, Canada and Germany are currently also having a discussion about AI-related cybersecurity risks.