Anthropic's Claude Mythos: Fear Marketing or Real Risk?

Anthropic recently announced Claude Mythos, a model it claims surpasses human experts at finding cybersecurity bugs. In an early April blog post, the company warned of severe fallout for economies, public safety, and national security if similar technology lands in wrong hands. Some observers even predicted mass replacement of devices, from laptops to Wi-Fi microwaves.

But security experts doubt the claims, and critics see a pattern: AI companies amplifying existential risk narratives to distract from real-world damage, inflate stock prices, and position themselves as the only responsible stewards. Shannon Vallor, ethics professor at University of Edinburgh, argues that framing AI as “supernatural in danger” makes regulators feel powerless—leaving only the companies themselves as guardians.

This isn't new. In 2019, Anthropic CEO Dario Amodei, then at OpenAI, helped declare GPT-2 too dangerous to release, citing malicious applications. Months later, OpenAI released it anyway, with Sam Altman later calling those fears “misplaced.” Altman recently criticized Anthropic's “fear-based marketing,” yet his own playbook included similar warnings that AI “will probably most likely lead to the end of the world, but in the meantime, there'll be great companies.”

Anthropic's spokesperson declined to address the article's points but shared supporting blog posts from other organizations. The pattern suggests fear is a deliberate strategy—whether or not Mythos is as dangerous as claimed.