Prometheus rising: Anthropic's frightening new baby is here

Anthropic's Claude Mythos AI finds critical vulnerabilities at a superhuman level, raising global security fears despite "safety-first" claims.

Anthropic is a company that's having it every which way.

Prudent: it warns about the dangers of the AI products that can be produced. Principled: it tells the Trump administration it will not partake in creating AI software that aids mass surveillance -- a move that earned it an order of excommunication as a "supply chain risk".

Despite these high-mast principles, the company now reveals it has produced something bound to concern governments the world over.

The new Claude Mythos Preview model has been advertised as the "best-aligned model that we have released to date by a significant margin".