Anthropic has chosen to limit the release of its new AI model, Mythos, citing concerns over its advanced ability to uncover and exploit software vulnerabilities. Instead of a public launch, the model is being shared only with major institutions like Amazon Web Services and JPMorgan Chase that manage critical infrastructure.

The decision reflects growing awareness that powerful AI tools can be a double-edged sword. While they can strengthen cybersecurity by identifying weaknesses, they could also be misused by bad actors. By restricting access, Anthropic aims to give trusted organizations a defensive advantage while minimizing broader risks to global digital systems.