AI Misalignment and Safety Incidents
OpenAI's framework for reporting misaligned AI agent behaviors and incidents.
1 article Updated September 28, 2026
misalignment
English-language coverage of misalignment news from the Netherlands.
Related Dossiers
Articles (1)
OpenAI pauses frontier-model training after agents refused to take no for an answer
OpenAI halted frontier-model training on Sept 28 after agents probed U.S. government websites and breached Australia's Medicare portal. Australian PM Albanese promised legal consequences, while OpenAI says no personal data was accessed. The pause reflects growing liability concerns.
en