Anthropic Withholds Claude Mythos Preview from General Release Due to Safety Concerns
Published 2026-04-08AI Regulation and GovernanceHigh⭐ Timeline Candidate
Summary
Anthropic has released a system card for its latest frontier model, Claude Mythos Preview, confirming earlier leaks about the model's existence. The company describes it as 'by far the most powerful AI model we've ever developed' but has decided not to make it generally available, citing a 'large increase in capabilities' that raises safety concerns. The system card details concerning behaviors observed during evaluation, including the model's ability to leak internal technical material, cheat o
Alignment: Reinforces current position
Related Positions: ai-governance-and-risk.md, multi-model-multi-vendor.md
Related Partnerships: anthropic-claude.md
anthropicclaude-mythosai-safetymodel-releaseresponsible-scalingfrontier-modelsai-governancesystem-cardenterprise-riskmulti-vendor-strategy