EU welcomes Anthropic's move to slow the release of powerful new AI tool (politico.eu)

• Claude Mythos demonstrates high-proficiency in autonomous cyber-attack vectors.
• Voluntary release delay sets a new industry benchmark for safety transparency.
• EU policymakers view the move as a validation of emerging AI liability and security standards.

"Anthropic has delayed the deployment of its latest model, Claude Mythos, after internal safety testing revealed it possesses advanced capabilities to autonomously identify and exploit software vulnerabilities, outperforming most human hackers. The EU has lauded this proactive caution, marking a significant development in AI safety protocols and alignment with stringent risk-mitigation frameworks for foundation models."