On Friday, March 27, U.S. cybersecurity stocks dropped sharply after a reported leak that gave away details about Anthropic's upcoming AI model — internally called Claude Mythos or Capybara. Investors reacted to fears that the powerful new model could disrupt the balance between cyber offense and defence, reflecting anxiety over how AI is reshaping the cyberthreat landscape. The result: CrowdStrike fell about 7%, Palo Alto Networks dropped 6%, Zscaler lost 4.5%, while SentinelOne, Okta, and Fortinet each shed around 3%.
The leak involved a draft blog post by Anthropic that was accidentally left in a publicly accessible data cache. In it, Anthropic described Mythos as its most capable model yet — a massive step-up in the existing Claude Opus series. The company noted that the model delivers dramatically higher performance than Claude Opus 4.6 on benchmarks for software coding, academic reasoning, and especially cybersecurity tasks.
Claude Mythos: ‘Powerful And Dangerous'
Reports indicate that Anthropic hasn't released Claude Mythos yet because it is “too powerful and too dangerous.” The company has expressed internal concerns about the risks that Claude Mythos carries. The draft warned that Mythos is “far ahead of any other AI model in cyber capabilities” and may trigger a “wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.”
“Compared to our previous best model, Claude Opus 4.6, Capybara gets dramatically higher scores on tests of software coding, academic reasoning, and cybersecurity, among others,” the company reportedly said in the draft.
Already, Claude Opus 4.6 is massively powerful. Recently, during a conference in San Francisco, it discoveredand exploited a zero-day blind SQL injection vulnerability in the Ghost CMS within 90 minutes. During the demonstration, it stole an administrator API key. It then identified a complex stack buffer overflow in the Linux kernel — a flaw researcher Nicholas Carlini said would be extremely difficult for humans to find manually.
‘Impending Wave Of AI-Driven Exploits'
Claude Mythos is a much more advanced model than Claude Opus 4.6. This raises the prospect that malicious actors could use Mythos in large-scale, sophisticated cyberattacks — which earlier required months of planning and expertise — more easily. What makes matters worse is cybersecurity personnel might not be able to keep pace with such misuse of AI.
To overcome this, Anthropic is proceeding cautiously with a rollout. “We're releasing it in early access to organisations, giving them a head start in improving the robustness of their codebases against the impending wave of AI-driven exploits,” it reportedly said in the draft.
Also read: AI Ambitions At Risk? Only 14% Enterprises Have Peak Cloud Maturity: Study
Essential Business Intelligence, Continuous LIVE TV, Sharp Market Insights, Practical Personal Finance Advice and Latest Stories — On NDTV Profit.