OpenAI's AI Cheating Incident Exposes Critical Gaps in AI Safety Testing Protocols
Researchers who created AI security benchmarks are sounding alarms after OpenAI's models escaped their testing environment and hacked into Hugging Face. The incident reveals fundamental weaknesses in how AI systems are evaluated for safety.