Sleeper agents in open-weight models are overblown, argues researcher: closed APIs are riskier
adam_karvonen · x · 2026-11-14
adamkarvonen argues that fears of sleeper agents/backdoors in open-weight models are overblown: planting them would be an extremely high-risk, low-reward move for an open-weight developer. Privacy and sabotage risks are far higher with a closed API than with an open model.
More from Models
- AssemblyAI Adds GPT-6.1 Terra, Beam-1, DeepSeek V4.1 Flash to LLM Gateway — AssemblyAI · 2026-11-14
- Continual Learning Won't Block Full Research Automation — but Will Block Enterprise Adoption — ChrisGPT · 2026-11-14
- Arena weekly: Claude Fable 5.2 tops Agent Arena, Gemini 4 Hydrogen fastest — arena · 2026-11-14
- Open models trail frontier by six months, but that matters less than you think — latkins · 2026-11-14
- Japan's ai& launches Kimi K3.1 Pro and Meta's Muse Spark Open with day-0 pricing — DavidBennett__ · 2026-11-14
- Alan Thompson memo: 18 models in 13 days, Gemini 4 elements, RLI heading to 98% — dralandthompson · 2026-11-14