Two months ago, the Trump administration publicly branded Anthropic a "RADICAL LEFT, WOKE COMPANY" and a menace to national security. That hostility may be softening - and The Verge reports the reason is a new cybersecurity AI model called Claude Mythos Preview.
The tension carried real business risk for Anthropic. Government and enterprise contracts matter enormously to an AI lab burning hundreds of millions per year on model training, and a White House that views your company as an ideological enemy is not a customer. A cybersecurity model changes the pitch - suddenly you're not a woke tech company, you're a defense asset.
Claude Mythos Preview is designed for security teams rather than general users, aimed at helping analysts understand attack patterns, review code for vulnerabilities, and strengthen defenses. It positions Claude as infrastructure for national security work, which is exactly the kind of framing that plays well in current Washington. Whether the model's actual capabilities back up that positioning is a separate question Anthropic hasn't fully answered yet - no detailed benchmarks have been published.
Anthrop isn't alone in this play. OpenAI launched its own cybersecurity-focused model, GPT-5.4-Cyber, the same week. Both companies are chasing government and enterprise security contracts, and both are clearly betting that specialized models beat general-purpose tools in high-stakes domains where a wrong answer has consequences.
For security professionals already using Claude, Mythos Preview is worth tracking. But "preview" is doing real work in that name - this is an early release, and the gap between a model that helps with security work and one that's actually trusted for it is wide.