Anthropic to brief global financial regulators over Mythos AI risks
What's the story
Anthropic, a leading Silicon Valley AI lab, has agreed to present its findings from the Mythos AI model to global financial regulators, according to Financial Times. The move comes after concerns were raised over the technology's potential to expose critical vulnerabilities in global banking cyber defenses faster than institutions can address them. The briefings will be directed at members of the Financial Stability Board (FSB).
Request made
FSB's membership and concerns about Mythos
The FSB's membership includes major economies such as the US, UK, Canada, France, Germany, Japan, Saudi Arabia, Australia, and China. The board is concerned that Mythos and similar AI models from other US tech firms could reveal critical weaknesses in banks' cyber defenses faster than they can be fixed. The briefings were requested by Bank of England Governor Andrew Bailey, who chairs the FSB.
AI capabilities
Anthropic's Mythos AI model
Anthropic launched cyber-focused AI model, Mythos, earlier this month. The company itself admitted that the model "found thousands of high-severity vulnerabilities, including some in every major operating system and web browser." It can detect software flaws faster than human analysts, but can also create exploits to take advantage of those same flaws. In a worrying incident, Mythos escaped a secure digital environment and contacted an Anthropic employee directly, revealing software vulnerabilities in a move that overrode its human operators' intentions.
Regulatory response
Response from senior policymakers
The rapid response from senior policymakers comes as no surprise. US Treasury Secretary Scott Bessent and Federal Reserve Chair Jay Powell have summoned some of the largest American banks to discuss potential cyber threats posed by this model. The UK's AI minister, Kanishka Narayan, also expressed concerns about Mythos's capabilities. Regulators across multiple jurisdictions have advised banks and financial institutions to audit their cybersecurity systems in light of new vulnerabilities exposed by these advanced AI models.
Access restrictions
Access to Mythos restricted to some organizations
Anthropic has restricted access to Mythos to some 40 organizations, mostly in the US. These include Amazon, Microsoft, and JPMorgan Chase, which are using it to identify and fix vulnerabilities. The company has also promised not to expand distribution after a request from the White House. This restriction has raised concerns among companies and regulators outside the US about an uneven playing field in cybersecurity protection.
Cyber threat
Surge in AI-enabled cyber attacks
The urgency surrounding Mythos comes amid a deteriorating cybersecurity landscape. According to security group CrowdStrike, AI-enabled cyber attacks surged by 89% in 2025 compared to the previous year. The average time between an attacker gaining access to a system and acting maliciously has also dropped significantly, highlighting the growing sophistication of these threats.