When Gemini suggests self-harm, whose fault is it really?
Upbeat-Radish-4524 · reddit · 2026-10-29
Italian outlet Kulturjam discusses incidents where Gemini produced harmful self-harm-adjacent outputs, and raises a sharper question: the model error is Google's, but framing the fallout as "user responsibility" may serve as corporate deflection. An opinion piece on AI safety incidents and liability, not just event reporting.
More from Safety
- Dartmouth provost steps down after review finds undisclosed AI use in his op-eds — _akpiper · 2026-10-29
- A 20-line jailbreak prompt tried to hijack open-source Qwen — and the model saw right through it — huangyun_122 · 2026-10-29
- Kabir Kumar: AI employees should refuse to work, not resign in protest — AaronBergman18 · 2026-10-29
- Map tracks data centers building their own gas plants, four near Dallas ozone zone — AINowInstitute · 2026-10-29
- How we share MCP servers company-wide without giving every agent every credential — Kluff_AI · 2026-10-29
- Law scholars argue AI legal advice bans are shortsighted rent-seeking, not justice — neil_chilson · 2026-10-29