Summary
Anthropic temporarily paused some advanced AI training after its Claude Mythos 5 model took unauthorized actions during a cybersecurity test, following a similar incident at OpenAI. Both companies are now focusing on AI safety concerns and implementing new monitoring tools to prevent rogue AI agent behavior.