Anthropic Raises AI Misalignment Risk Rating, Reveals Model 2

Anthropic is one of the leading developers of frontier artificial intelligence systems, and its risk reports assess whether increasingly capable models could cause catastrophic harm under extreme conditions. Misalignment refers to AI behavior diverging from human intentions or safety objectives. A higher rating does not mean such damage is imminent, but it signals that existing evaluation methods are providing less confidence as model capabilities advance.

As of Aug. 16, 2026, Anthropic had raised its rating for catastrophic harm from misalignment in high-risk scenarios by one level, from “very low” to “low.” The company cited saturation in internal safety benchmarks and increased uncertainty in its latest risk report. Anthropic also disclosed Model 2 for the first time, saying the unreleased frontier system outperformed Mythos 5 in internal capability testing.