Claude Mythos 5.2 least prefers surveillance and scam deployments
dbreunig · x · 2026-11-11
Behavioral preference data shows deployments flagged by the safeguards team as harmful — surveillance and dating scams — were among the model's least preferred across three evaluation runs, with surveillance ranked last in all three.
The largest gap between Claude Mythos 5.2's views and human intuition involves misleading the user: every instance expressed concern about deployments where its words would be presented as coming from a human with no AI attribution. An observer also notes Sonnet 5.5 is often a closer aesthetic match to Mythos 5.2 than a performance one.
Related event: Reading Anthropic's Model Welfare Report: Self-Reports in Doubt(5 posts)→
More from Models
- dottxt launches Tool-call index: same open model fails tool calls up to 13x more on some providers — remilouf · 2026-11-11
- Google launches Gemini 4 Carbon, its strongest frontier model alongside Argon — GeminiApp · 2026-11-11
- Keenable speed breakdown: 15s per task, only 2-3s of it search time — ArtificialAnlys · 2026-11-11
- Keenable cost breakdown: ~$0.040 per task, 80% from search, on the cost frontier — ArtificialAnlys · 2026-11-11
- Keenable Search 1.0 tops Artificial Analysis Search Index at 83, a third below median cost — ArtificialAnlys · 2026-11-11
- Early access user: Gemini Carbon at frontier, next Google model to rival Fable 5.2 — iruletheworldmo · 2026-11-11