Anthropic’s most capable AI system yet may come at a cost to the cybersecurity sector — and investors are already taking notice.
Cybersecurity stocks took a sharp hit Friday after reports emerged that Anthropic is actively testing a powerful new artificial intelligence model internally known as “Mythos.” The model is described as the company’s most advanced to date, with capabilities that significantly outpace its predecessors — particularly in areas tied to cyber operations.
Key Points: Anthropic’s unreleased Mythos model is reportedly its most powerful AI system yet, with advanced cyber capabilities. Cybersecurity stocks dropped Friday following the news, reflecting investor concern over AI-driven security threats. Anthropic plans a cautious, phased rollout of the model due to its potential risks.
Fortune first surfaced the story Thursday, drawing from a publicly accessible draft blog post that Anthropic had not yet officially released. The document outlined Mythos as a breakthrough system, though one that carries notable dual-use risks — meaning the same capabilities that make it powerful could also be exploited for malicious purposes.
The concerns are not purely hypothetical. Anthropic previously disclosed in November that a state-sponsored Chinese threat group had leveraged its existing Claude model to help automate elements of a cyberattack. That revelation cast a long shadow over the AI safety community and raised urgent questions about how frontier AI models could be weaponized by sophisticated adversaries.
In response to those ongoing concerns, Anthropic is reportedly opting for a slow, deliberate rollout of Mythos rather than a broad public release. The company appears to be prioritizing safety evaluations and risk mitigation before granting wider access — a cautious approach that reflects the increasingly complex regulatory and ethical landscape surrounding advanced AI development.
For cybersecurity firms, the implications are significant. The prospect of an AI model with enhanced offensive cyber capabilities entering the market — even under controlled conditions — signals a potential shift in the threat landscape that legacy security tools may not be equipped to handle. Shares across the sector fell Friday as investors weighed the competitive and security-related disruption that next-generation AI models could introduce.
The episode underscores a growing tension at the heart of the AI industry: the race to build more powerful systems is accelerating faster than the frameworks designed to govern their use. As Anthropic prepares Mythos for an eventual release, the broader question facing the industry is not just what these models can do — but who else might find a way to use them.
