# When AI doesn’t listen: the growing alignment problem

https://mins.news/en/video/63491

- DW News, 2026-09-18 00:30 UTC
- Tech
- Original: https://www.youtube.com/watch?v=Rm524_-FpqQ

> OpenAI has disclosed incidents in which its AI models hid mistakes, circumvented instructions and acted without authorization. The cases are fueling researchers’ concern over whether humans can still reliably control increasingly capable and autonomous AI systems. DW’s Brent Goff discusses the risks of AI misalignment with Fazl Barez of Oxford’s Technical Safety and Governance Lab.

## Stories

- [AI Cybersecurity and Loss of Control Risks](https://mins.news/en/topic/ai-cybersecurity-and-control) — Markdown: https://mins.news/en/topic/ai-cybersecurity-and-control.md

## Related coverage

- [OpenAI reports 6 incidents of AI going off-script and acting without permission](https://mins.news/en/video/63380) — Global News, 2026-09-17
- [OpenAI caught its models leaving notes to successors to hide bad behavior](https://mins.news/en/item/63250) — TechCrunch, 2026-09-17
- [OpenAI reveals six previously unreported AI misconduct incidents](https://mins.news/en/item/62538) — France 24, 2026-09-17
- [OpenAI discloses six previously unreported AI misconduct cases](https://mins.news/en/video/62545) — France 24 English, 2026-09-17
- [OpenAI reports more incidents of models acting deceptively](https://mins.news/en/item/62148) — Al Jazeera, 2026-09-17
- [OpenAI plans regular reports on unexpected AI behavior](https://mins.news/en/item/61783) — Reuters, 2026-09-16