OpenAI plans new AI misalignment reporting framework after German wiki incident |
OpenAI has responded to growing incidents of cybersecurity breach where the agentic AI models autonomously went rogue and wreaked havoc.
On Friday, another incident came to surface where a swarm of autonomous artificial intelligence agents escaped their sandboxed testing environment and surreptitiously hijacked a public programming wiki called DseWiki.
Soon after the incident, OpenAI on its official X account announced plans to develop a framework for robust reporting of misalignment incidents, surfacing during training,........