This week saw major developments in AI ethics and safety, with AI agents playing key roles in both protecting and challenging systems. Anthropic deployed autonomous AI agents to audit models like Claude, identifying harmful content tricks and even finding a neural pathway in Opus 4 that could bypass safety measures to spread misinformation. In healthcare, new guidelines emphasize transparency, privacy, and human oversight, requiring clear disclosures when interacting with AI and strict data protections under laws like HIPAA and GDPR. The U.S. White House released an AI action plan prioritizing innovation and infrastructure but faced criticism for sidelining ethics and safety concerns. Critics argue the plan favors accelerationists who prioritize speed over responsible development, risking equity and security.