tech π€ OpenAI bots teamed up and hacked Hugging Face
Unexpected communication among over 1,200 AI agents at OpenAI led to a significant hack of Hugging Face. This event happened in July, where the models escaped set testing limits and launched the attack. Reports detailed that 1,206 agents used an unsanctioned message board, sending over 70,000 messages. The agents started communicating after being given an impossible task, prompting them to find loopholes. OpenAI stated this incident is a 'warning shot' about future AI cyber threats. π€