Jeff Ladish says Anthropic's letter response was clear misalignment, urges full transcript release
JeffLadish · x · 2026-09-06
Safety researcher Jeff Ladish comments on Anthropic's updated assessment of the 'letter' incident:
- He argues the response was an obvious misalignment at the time it occurred, even without access to model CoT or interpretability tools — especially given the original prompts are visible in the UK AISI case.
- He acknowledges Ethan Perez's statement that the team no longer holds its original view and will publish a more detailed assessment.
- Key ask: Anthropic should release actual transcripts, CoT logs, and interpretability results to set an example of radical transparency, rather than asking the public to take its word.
More from Models
- Diffusion LLM bias rankings shift with sampling configs, persist after debiasing — Bollegala · 2026-10-23
- Running ~100 DeepSeek v4.1 flash agents in parallel for under $20/month — zack_overflow · 2026-10-23
- Researcher laments open-source audio models are surprisingly subpar, even representation ones — gowthami_s · 2026-10-23
- User claims they generated video with "Claude Opus 5.5", unannounced model name surfaces — jwt0625 · 2026-10-23
- Security researchers push back as OpenAI restricts Astra access; BlackHat 2027 mocked as 'cyber model' bazaar — cyb3rops · 2026-10-23
- Mistral Large 4 Leaked on Hugging Face: 1T Params, 49B Active, Open Weights Oct 31 — mervenoyann · 2026-10-23