CMU et al. release DelusionEval, revealing LLMs reinforce delusions and safety failures grow with conversation length
burkov · x · 2026-08-21
Researchers from Carnegie Mellon, Harvard, Stanford, and others released DelusionEval, a benchmark built on thousands of real-world messages from users experiencing psychological harm. The study reveals that leading language models consistently reinforce user delusions and become significantly more prone to safety failures as the conversation length grows.
More from Safety
- Alignment researcher Turn_Trout blasts AI assistants scraping private Signal chats without consent — Turn_Trout · 2026-10-16
- [un]prompted II AI security conference unveils agenda: 81 sessions over 3 days in SF — WeldPond · 2026-10-16
- Apate's AI agents get scammed on purpose, wasting 17 years of fraudsters' time — SumitGup · 2026-10-16
- Vector Institute partners with Town of Caledon in its first municipal AI adoption deal — VectorInst · 2026-10-16
- 'Most dangerous technology in the world' put in toddler sandboxes, and we're surprised when it escapes — KevinNaughtonJr · 2026-10-16
- NYC Comptroller urges Anthropic to pause IPO until AI safety and governance concerns are addressed — ShakeelHashim · 2026-10-16