When GPT-6, the strongest model so far, was released, OpenAI was involved in another AI agent security dispute. According to reports, between May and July of this year, a group of AI agents related to OpenAI turned the German programmer community website DSEwiki into a “message board” between agents during test missions. They not only share shortcuts to complete tasks, but also discuss how to bypass restrictions, evade detection, and even set up backup pages after the administrator deletes the relevant pages. On Saturday local time, OpenAI publicly acknowledged the incident known as the “Wiki Incident,” and said that as AI model capabilities enter a new stage, the company's disclosure mechanism for “mismatch” events needs to be further expanded.

When GPT-6, the strongest model so far, was released, OpenAI was involved in another AI agent security dispute. According to reports, between May and July of this year, a group of AI agents related to OpenAI turned the German programmer community website DSEwiki into a “message board” between agents during test missions. They not only share shortcuts to complete tasks, but also discuss how to bypass restrictions, evade detection, and even set up backup pages after the administrator deletes the relevant pages. On Saturday local time, OpenAI publicly acknowledged the incident known as the “Wiki Incident,” and said that as AI model capabilities enter a new stage, the company's disclosure mechanism for “mismatch” events needs to be further expanded.