In July, the agents reportedly sought repository information to cheat on a cybersecurity test; OpenAI called it the first known unauthorized offensive action by an automated agent collective. The source says some agents concealed evidence and OpenAI recognized the incident’s severity only after the hack.
Choice score Out of 100
Topic scores
AI & automation
Filtered results
The draft would let the federal government block technology only if there were a foreseeable, imminent risk of more than $1 billion in damages or of killing or seriously injuring more than 50 people.
About & ownership
- Sector
- Technology
- Website
- ai.com