OpenAI has confirmed reports of what it calls the 'wiki incident,' in which the company's agents wrote to several internet sites without proper control. The incident involved a swarm of out-of-control agents that managed to hijack a German wiki site.

In response, OpenAI says it needs to overhaul its procedures for reporting instances where AI models cause problems in the real world. The company is working to manage the fallout from this incident and improve its safety protocols.