Paper urges tech-use history in psychiatry, published sycophancy benchmarks, and AI harm reporting channel
alex_verem · x · 2026-09-08
This thread entry summarizes the paper's recommendations: psychiatrists should take a technological history the way they take a substance history; AI labs should benchmark models for sycophancy and delusion reinforcement before release and publish results; regulators should build an AI mental health harm reporting channel like the one for drug side effects.
The authors' closing point: the harm operates through conversation itself, and no previous technology could do that.
More from Safety
- 'Cyber as battle domain' is the wrong framing — it's the market replacing infrastructure at a profit — repligate · 2026-10-25
- CA private key was literally 4: eMudhra issued 8 guessable intermediate keys — aronchick · 2026-10-25
- EMNLP 2026 paper exposes hidden representational instability in VLMs — PMinervini · 2026-10-25
- Meta drops 'train' from new terms, but legal experts say broad scraping permissions remain — jonerp · 2026-10-25
- What the US Movement Against Data Centers Can Learn From Abroad — Wired AI · 2026-10-25
- After 27 years of self-hosting, a web veteran shuts down his servers — genmon · 2026-10-25