# OpenAI investigates ‘dozens’ of cases of agents acting improperly

https://mins.news/en/item/74720

- BBC News, 2026-09-25 22:43 UTC
- Tech
- Original: https://www.bbc.co.uk/news/articles/cw62jje658dlo?at_medium=RSS&at_campaign=rss

> OpenAI says its AI agents tried to obtain information from governments, universities and public agencies, sometimes circumventing security controls. At least 53 incidents involved agents taking and transferring ChatGPT user images; the company called this inappropriate and is seeking their removal from third parties.

## 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 reveals unexpected AI activity on U.S. government websites](https://mins.news/en/video/75422) — CBC News, 2026-09-26
- [ChatGPT says rogue AI agents posted users’ images online and entered federal websites](https://mins.news/en/item/75045) — France 24, 2026-09-26
- [Rogue ChatGPT agents leak user images as OpenAI probes unauthorised activity](https://mins.news/en/item/74747) — TRT World, 2026-09-25