RESEARCH Security Researchers Used Anthropic’s Claude to Break Into OpenAI in Under 72 Hours 8/10 2 min read 4 hours ago