A single, leaked internal document from AI lab Anthropic wiped roughly $45 billion from the value of leading cybersecurity firms on March 27, 2026. The cause was a draft blog post detailing 'Claude Mythos,' a new AI model the company described as a 'step change' in capability that presents 'unprecedented cybersecurity risks.'

The market's reaction was immediate and severe. The iShares Cybersecurity ETF fell over 3%. Shares of CrowdStrike, Palo Alto Networks, and Zscaler dropped between 5% and 7%. Tenable, a vulnerability management specialist, plunged as much as 11%. This was not a fleeting glitch; it was a deliberate, sector-wide reassessment of how generative AI threatens established security businesses.

According to the leak, Mythos demonstrates an ability to find novel software vulnerabilities and chain together complex attacks at a speed that dwarfs human-led teams. This isn't theoretical. In late 2025, Anthropic disclosed that a Chinese hacking group used a prior model, Claude Code, to autonomously run nearly 90% of a major espionage campaign.

For security leaders, the path forward is clear, if difficult. Threat models must formally account for AI-augmented adversaries. Defensive tools must themselves leverage AI to keep pace, potentially using similar frontier models for internal testing. Supply chains for AI developer tools require stricter governance.

The selloff reflects a harsh new reality: core functions of the cybersecurity stack, particularly in vulnerability management, are vulnerable to being commoditized by advanced AI. Companies that integrate these capabilities, like CrowdStrike and Palo Alto Networks, may adapt. Those that cannot keep up face an existential challenge. The race is now AI against AI, and the defensive side has just been put on notice.

Source: Webpronews