This data leak confirms what internal documents already suggested: Anthropic is testing Claude Mythos its most capable model to date and a clear step change beyond Opus 4.6 with select early-access customers.
The company’s own assessment is particularly notable: the model demonstrates unprecedented performance in cybersecurity benchmarks, raising risks that could enable future systems to exploit vulnerabilities far faster than defenders can respond.
Anthropic’s deliberate, phased rollout prioritizing cyber defenders reflects the gravity of these dual-use implications. As frontier AI advances accelerate, this episode underscores the growing need for coordinated safety measures across the industry.
BREAKING: Anthropic data leak reveals the existence of “Claude Mythos,” a new AI model that reportedly presents unprecedented cybersecurity risks.