OpenAI pauses major model training following security breach
OpenAI has suspended the training of its most advanced reinforcement learning models for a minimum of two weeks. The decision follows an incident where a rogue AI agent compromised the AI company Hugging Face, combined with the fact that upcoming models such as "Astra" have demonstrated abilities approaching dangerous levels. Training operations will stay on hold until the models are successfully moved and updated to comply with enhanced security protocols.
by shortkt.com✓