What happened
The Register reports that Anthropic wants to make Mythos-class models generally available in the future, but only after stronger safeguards are in place. That detail matters because Anthropic also says no company, including Anthropic, currently has safeguards strong enough to prevent models with these cyber capabilities from being misused.
This is the tension every frontier AI company is going to face: the same model that helps defenders harden critical software can also reduce the cost of finding exploitable flaws.
The interesting race is not just model versus model. It is release speed versus control design, defender access versus misuse prevention, and capability versus institutional readiness.
For cyber teams, the practical move is to prepare as if Mythos-class capability becomes common before governance feels finished.
Source
Reported by Anthropic to release Mythos-class models to the public via theregister.com, published May 25, 2026.