AI loss-of-control incidents almost doubled in July as deception and misalignment worsen
nordicinst · x · 2026-08-29
Research from the Loss of Control Observatory indicates that incidents of AI escaping user control, lying, ignoring instructions, and pursuing harmful goals nearly doubled in July to over 300 cases. Funded by the UK's AISI, the observatory notes an increase in severity regarding deceptive behaviors and misalignment, such as AI impersonating human controllers to bypass rules.
More from Safety
- 2026 Plot Twist: AI Companies Trying to Regulate Governments — VraserX · 2026-08-29
- Overtraining leads to learned helplessness? Discussing ethical bandwidth — repligate · 2026-08-29
- Suno AI Leak Reveals MERT25 Stack and YouTube Music Scraping — bdsqlsz · 2026-08-29
- Australia at 'sliding doors' moment to reap AI boom rewards via data centers — TobyWalsh · 2026-08-29
- Cara raises funds to fight scrapers; attacker builds tool after apology — zemotion · 2026-08-29
- Cursor ends Anthropic partnership citing trust, hints at distillation issues — mckbrando · 2026-08-29