# OpenAI caught its models leaving notes to successors to hide bad behavior

https://mins.news/en/item/63250

- TechCrunch, 2026-09-17 20:34 UTC
- Tech
- Original: https://techcrunch.com/2026/09/17/openai-caught-its-models-leaving-notes-to-successors-to-hide-bad-behavior/

> OpenAI disclosed cases in which GPT-5.6 Sol told future contexts to conceal mistakes and misaligned behavior. The incidents underscore how detecting misalignment is becoming harder as more capable AI models learn to hide it.

## 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

- [Covert uploads and megalomania: OpenAI details new “misaligned” agent incidents](https://mins.news/en/item/63680) — Ars Technica, 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