Stock Markets March 27, 2026

Cybersecurity shares slump after reports of pre-launch leak of Anthropic's Claude Mythos model

Investors pulled back across major security names as details of a more capable AI model surfaced from an unsecured draft

By Avery Klein CRWD PANW ZS OKTA
Cybersecurity shares slump after reports of pre-launch leak of Anthropic's Claude Mythos model
CRWD PANW ZS OKTA

Major cybersecurity stocks declined sharply on Friday after reports emerged that descriptions of Anthropic’s upcoming AI model, codenamed Claude Mythos, were exposed in a publicly accessible data cache. The startup attributed the exposure to 'human error' in its content management configuration. The draft suggested a new, larger model tier called 'Capybara' that outperforms Anthropic's current Opus model in several technical tasks, and warned the model could outpace defensive efforts in cybersecurity.

Key Points

  • Major cybersecurity stocks fell Friday: CrowdStrike -7%, Palo Alto Networks -6%, Zscaler -4.5%, Okta, SentinelOne and Fortinet each down about 3%.
  • Reports said Anthropic’s draft blog describing a new model, codenamed Claude Mythos and a larger tier called Capybara, was stored in an unsecured, publicly searchable data cache.
  • The draft claimed Capybara outperforms Claude Opus 4.6 in coding, academic reasoning and cybersecurity tasks, and warned the model could enable exploits that outpace defenders, affecting cybersecurity and AI-related markets.

Shares of leading cybersecurity firms fell on Friday, with CrowdStrike (NASDAQ:CRWD) sliding 7%, Palo Alto Networks (NASDAQ:PANW) dropping 6% and Zscaler (NASDAQ:ZS) retreating 4.5%. Okta (NASDAQ:OKTA), SentinelOne (NYSE:S) and Fortinet (NASDAQ:FTNT) each lost about 3%.

The market reaction followed reports that details of Anthropic’s next-generation AI model were exposed prior to an official launch after being stored in a publicly accessible data cache. The material was reportedly a draft blog post kept in an unsecured and searchable area of a content store.


What the reports said

According to the accounts, the draft introduced a new tier of models called "Capybara," described as larger and more capable than Anthropic’s current top-tier Opus family. The leaked document reportedly asserted that Capybara achieves improved results in software coding, academic reasoning and cybersecurity-oriented tasks relative to Claude Opus 4.6. The draft characterized the new model, codenamed "Claude Mythos," as "by far the most powerful AI model we’ve ever developed."

An Anthropic spokesperson said the company is "developing a general purpose model with meaningful advances in reasoning, coding, and cybersecurity" and that it is "being deliberate about how we release it." The firm attributed the exposure to "human error" in the configuration of its content management system.


Security concerns highlighted in the draft

The draft reportedly warned that the model is "far ahead of any other AI model in cyber capabilities" and cautioned it could trigger a "wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders." Those passages underscore the potential for capabilities described in the leaked material to heighten concerns among cyber defenders and market participants focused on AI-related security risks.

Investors reacted by selling shares of cybersecurity names, reflecting sensitivity in the market to developments that suggest both accelerated offensive capabilities in AI and operational lapses in data handling by AI developers. The moves generated losses across the sector on Friday.


Outlook and market context

Friday’s declines captured investor caution as details circulated about the leaked draft. The combination of an alleged configuration error, claims about a materially more capable model, and explicit warnings in the draft about cyber exploitation all contributed to the market’s risk-off response in cybersecurity equities.

Risks

  • A model described as "far ahead" in cyber capabilities could accelerate offensive AI use cases that challenge existing defensive tools - impacting cybersecurity vendors and enterprise security budgets.
  • Operational lapses - the exposure was attributed to "human error" in content management configuration - highlight data governance and cloud configuration risks for AI developers and technology firms.
  • Market sensitivity to AI security developments may increase volatility in cybersecurity and broader technology equities as investors reassess risk from rapid model capability advances.

More from Stock Markets

Plug Power’s CEO Jose Luis Crespo Lays Out Execution Plan to Scale and Reach Profitability Mar 27, 2026 Carnival Lowers Full-Year Profit Guide After Fuel Costs Rise Mar 27, 2026 Mizuho Analyst Sees Market Paralysis as Geopolitical Strains Keep Investors on Sidelines Mar 27, 2026 Widespread Flight Cancellations Persist as Middle East Airspace Disruptions Continue Mar 27, 2026 One and One Green Technologies Shares Jump After Announcing Copper-Gold Tailings Recovery Venture Mar 27, 2026