UK, US and Canadian Regulators Sound Alarm Over Anthropic's Claude Mythos AI Model

Financial authorities in multiple countries are conducting urgent consultations following the emergence of Anthropic's new AI model, Claude Mythos. The neural network discovered thousands of zero-day vulnerabilities in major operating systems and web browsers within just a few weeks, triggering alarm across global regulatory bodies, according to FT.

Emergency Meetings Across Three Continents

In the United Kingdom, representatives from the Bank of England, the Financial Conduct Authority (FCA), and HM Treasury have entered into discussions with the National Cyber Security Centre. Their goal is to jointly assess potential weaknesses in critical IT systems exposed by the model. Over the next two weeks, major banks, insurance companies, and exchanges will receive a cybersecurity risk briefing related to Claude Mythos.

In the United States, the scale of the threat prompted Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell to convene an emergency meeting with Wall Street executives. Attendees included Citigroup CEO Jane Fraser and Goldman Sachs CEO David Solomon. Officials warned that the launch of Mythos marks a new era in cybersecurity.

The Bank of Canada simultaneously organized a similar gathering with the country's leading financial institutions to evaluate systemic risks and strengthen the operational resilience of its critical financial sector.

Why This Matters

The Claude Mythos situation represents a qualitative leap in AI capabilities — and the threats that accompany them. If the model can indeed mass-identify zero-day vulnerabilities, it fundamentally shifts the balance of power in both cyber defense and potential cyberattacks. The coordinated response from regulators spanning the Fed to the Bank of England underscores the systemic nature of risks to global financial infrastructure.

Project Glasswing Replaces Public Launch

Mythos proved so powerful that Anthropic scrapped any plans for a public release. Instead, the company launched Project Glasswing — a controlled testing initiative involving major technology corporations: AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks.

Marketing Play or Genuine Threat?

TechCrunch noted that restricting access to Mythos could be part of a marketing strategy designed to generate buzz around the product.

Dan Lahav, head of AI startup Irregular, pointed out in March that while automated vulnerability discovery plays a significant role, the actual danger of any discovered flaw to malicious actors depends on numerous additional factors.

Anthropic claims Mythos surpasses its predecessor Opus in the breadth of vulnerability exploitation. However, startup Aisle reported it was able to replicate most of Mythos's results using other neural networks. According to Aisle, there is no single universal model for cybersecurity — tool selection is always dictated by the specific task at hand.

David Crawshaw, CEO of exe.dev, offered a sharper take, calling the situation a marketing ploy. He argued that top-tier models are now available only through corporate agreements and are no longer provided to smaller labs for testing. When Mythos eventually becomes broadly available, the market will simply gain another cutting-edge LLM exclusively for enterprise clients — a strategy that sustains funding inflows while restricting companies engaged in algorithm distillation.

Anthropic Gains Momentum as OpenAI Loses Its Edge

Anthropic has significantly expanded its user base in recent months, with revenue growing exponentially. According to TechCrunch, at the HumanX AI conference in San Francisco, thousands of AI professionals discussed the technology's potential, and Claude was mentioned most frequently as one of the best products on the market. Anthropic featured across numerous panel sessions, while ChatGPT was barely discussed.

One expert stated that OpenAI had "fallen out of the game." Many conference attendees noted that Sam Altman's company had lost a clear direction. The negative sentiment was amplified by a recent profile of the startup's CEO that characterized him as a dictator and a deceiver who abandoned the company's original noble mission.

Nevertheless, a financial analysis of both companies revealed that OpenAI and Anthropic remain "the fastest-growing enterprises in history." TechCrunch assessed that OpenAI has simply gained serious competitors — a normal development in any industry. The company has already responded by announcing a $100 tier that offers expanded access to its coding tool Codex.

Earlier in April, Anthropic released Claude Managed Agents — an environment for running complex, long-duration agentic tasks.