Anthropic Claims New AI Is Too Dangerous—Skeptics Call It a Hype Play

Anthropic claims its new AI model, Mythos, is too powerful for a public release, citing significant cybersecurity risks. But as reported by The Guardian, skeptics see a different motive: a calculated publicity campaign to generate hype and attract investment. The move positions Anthropic as both a technical leader and a responsible steward—a convenient narrative in the high-stakes AI funding race.

The company announced it had created the model but was withholding it out of an “overwhelming sense of responsibility,” according to The Guardian. This immediately frames Mythos as a uniquely potent technology. The announcement succeeded in capturing attention, with The Guardian noting that the US Treasury Secretary, Scott Bessent, summoned bank heads to discuss the model, and a UK MP also raised the issue. When government officials start holding meetings, a company's claims—or its marketing—have clearly landed.

A 'Preview' Release or a PR Campaign?

The story has a wrinkle. While The Guardian reports Anthropic is “not going to release it to the public,” Fast Company states the model was released in “preview.” This apparent contradiction points to a common industry strategy. A “preview” release typically means providing access to a select group of high-value partners, such as major corporations or cybersecurity firms, while maintaining the public narrative of cautious withholding. This allows Anthropic to monetize its cutting-edge model with enterprise clients while publicly performing safety.

Together, these reports suggest a dual-track strategy. The public message is one of restraint and safety, designed to build brand reputation and perhaps preempt regulatory scrutiny. The private track involves giving key partners early access, securing a commercial foothold. This isn't just about safety; it's about market positioning. In the battle against OpenAI and other deep-pocketed labs, demonstrating a capability so advanced you claim it must be contained is a powerful way to signal you're still in the race.

The Cybersecurity Debate

The entire premise of Anthropic's announcement rests on the model's alleged cybersecurity prowess. Fast Company surveyed cybersecurity professionals, revealing a sharp division on whether models like Mythos are a net positive. The core of the issue is whether they give an advantage to defenders or attackers. An AI that can autonomously discover and even exploit software vulnerabilities is a classic double-edged sword.

On one hand, security teams could use such a tool to proactively find and patch flaws in their own systems before malicious actors do. It could automate the tedious work of code review and penetration testing, dramatically improving defensive postures. On the other hand, the same tool could be used by attackers to generate novel exploits for zero-day vulnerabilities at an unprecedented scale. The experts are split because the outcome depends less on the technology itself and more on who masters it first. The pattern indicates that the debate over AI's role in security is far from settled, and Anthropic has placed itself directly at the center of it.