A significant leak of internal documents from AI firm Anthropic has revealed details of a powerful new model, triggering a sell-off in software and cybersecurity stocks early Friday. The incident underscores the market's sensitivity to AI's potential dual-use nature.

According to a Fortune report, roughly 3,000 internal assets, including draft blog posts, were inadvertently exposed online. The materials detail a model referred to internally as "Claude Mythos," described by the company as a "step change" and its most capable system to date. It is currently in limited testing.

The leaked drafts contained a stark warning: this advanced model could dramatically heighten cybersecurity risks by rapidly discovering and exploiting software vulnerabilities, potentially accelerating a digital arms race. Investors reacted immediately. Shares of major cybersecurity firms like Palo Alto Networks, CrowdStrike, and Fortinet fell between 4% and 6%. The iShares Expanded Tech-Software Sector ETF (IGV) dropped nearly 3%. The news also coincided with a slide in Bitcoin, pushing it back toward $66,000.

The documents also hint at a future model tier, codenamed "Capybara," positioned as larger and more intelligent than Anthropic's current top-tier model, Claude Opus. The leak presents a public relations challenge for Anthropic, which has built its reputation on a focus on AI safety, even as it races to develop more powerful systems.

Source: CoinDesk