AI safety debate: is doom-flavored risk communication from OpenAI and Anthropic actually working?
BronsonSchoen · x · 2026-09-09
BronsonSchoen and ericlim debated how frontier-lab risk messaging lands:
- Schoen argues that "we have to keep going because we're the responsible ones" is also bad; statements from OpenAI (Jakub's recent post) and Anthropic insiders mainly show that no one is on top of this, and something like "Pacing the Frontier" is needed
- He cites Evan Hubinger's public adaptation of an internal Anthropic document about communicating risk to colleagues
- ericlim counters that there's a misalignment between how risk is communicated and what follows — building something that could kill everyone while accelerating is incoherent
More from AGI Musings
- Anthropic pretraining researcher quits, accusing OpenAI and Anthropic of recklessly racing to self-improving superintelligence — TinfoilTricorn · 2026-09-10
- Researcher Blasts OpenAI Whistleblower Interview: Focus on AI Ethics, Not Alignment — examachine · 2026-09-10
- Researcher calls AI alignment 'safety theater': ethics, not alignment, is the real problem — examachine · 2026-09-10
- Don't buy the 'software engineering is doomed' narrative from AI labs eyeing IPOs — bendee983 · 2026-09-10
- Ex-DeepMind, now Anthropic researcher: no viable scientific plan for recursively self-improving AI risks — harris_edouard · 2026-09-10
- Hugging Face CEO: If AI Risk Is Real, Labs Must Openly Share Models — Cue the Sarcasm — Gradio · 2026-09-10