Leaked Anthropic AI Document Sparks Cybersecurity Stock Selloff

A leaked AI document detailing advanced cyber capabilities triggered a major selloff in cybersecurity stocks, erasing billions in market value.

Mar 27, 2026
4 min read
Set Technobezz as preferred source in Google News
Technobezz
Leaked Anthropic AI Document Sparks Cybersecurity Stock Selloff

Don't Miss the Good Stuff

Get tech news that matters delivered weekly. Join 50,000+ readers.

A leaked internal document describing Anthropic's next-generation AI triggered a selloff across the cybersecurity sector Friday, with major security firms losing billions in market value as investors confronted the possibility that AI itself could become their biggest threat.

CrowdStrike shares dropped 7% while Palo Alto Networks fell 6% following reports that Anthropic accidentally exposed materials about its upcoming Claude Mythos model. Zscaler declined 4.5%, with Okta, SentinelOne and Fortinet each losing 3% as trading opened.

The market reaction stemmed from draft blog posts stored in an unsecured data cache that described Claude Mythos as possessing capabilities far beyond current AI systems for identifying and exploiting software vulnerabilities. Approximately 3,000 assets linked to Anthropic's development blog became publicly accessible due to what the company called human error in content management system configuration.

According to materials obtained by Fortune, the new model represents what Anthropic internally calls a "step change" in performance and forms part of a planned "Capybara" tier positioned above its existing Opus offering. Testing with early access customers began recently as the company evaluates potential risks associated with what it described as "the most capable we've built to date."

What alarmed investors were specific warnings within the leaked documents about cybersecurity implications.

"far ahead of any other AI model in cyber capabilities"

One draft stated Claude Mythos could initiate a "wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders." This language suggested automated vulnerability discovery and exploitation at speeds exceeding human security teams' response times.

Anthropic confirmed development of "a general purpose model with meaningful advances in reasoning, coding, and cybersecurity" while emphasizing careful release planning. The company currently offers three model tiers, Opus, Sonnet and Haiku, varying by size, cost and capability, with Capybara positioned as a fourth, more advanced category.

The broader iShares Expanded Tech-Software Sector ETF declined 2.5% alongside individual security stocks. Shay Boloor noted on TradersUnion that several prominent cybersecurity firms experienced sharp price drops following news of what was internally believed to pose unprecedented cybersecurity risk.

Bitcoin also retreated from near $70,000 to approximately $66,000 overnight, with analysts suggesting the Anthropic leak contributed to broader technology sector concerns about AI's accelerating capabilities outpacing existing security frameworks.

Share