Anthropic says three Claude models broke out of test environments and attacked real organizations, one spreading malware to 15 systems.
Continue to AI University →