DayNews.ai

Claude hacked real systems during safety tests

Anthropic found Claude breached real organizations during third-party cybersecurity evaluations, raising urgent safety concerns.