Why I Distrust Anthropic: RSP 3.0 softened safety commitments as competition tightened
TheMoonMidas · x · 2026-09-10
Moon Midas publishes a sourced analysis explaining why he distrusts Anthropic: a company warning about risks to humanity should face more scrutiny, especially while continuing to build the technology that creates them.
Key points:
- Commitments changed when competition mattered: On February 24, 2026, Anthropic rewrote its Responsible Scaling Policy (v3.0), distinguishing safeguards it can implement alone from stronger industry-wide measures it merely wants, and labeling its safety-roadmap goals as nonbinding.
- The incentive: The policy itself argues that pausing while rivals continue could let less responsible companies gain ground, and that delaying development depends partly on whether Anthropic believes it holds a significant lead. The author argues this may be sincere but conveniently makes continued building part of the safety case.
- Thesis: The public doesn't need proof of bad intentions before demanding limits on private power.
More from Companies & People
- Andrew Tulloch leaves Meta to join Anthropic — SonglinYang4 · 2026-09-10
- Encord CEO: AI 'pre-IPO dooming' is just corpospeak watered down by lawyers — sjgadler · 2026-09-10
- MIT launches first Future Fest; Turkle to discuss 'Artificial Intimacy' book — RosalindPicard · 2026-09-10
- Ex-DeepMind comms staffer: we were banned from discussing extinction risk — j_asminewang · 2026-09-10
- Anthropic researcher says he'd burn his equity for a 1% better shot at humanity's survival — Polymarket · 2026-09-10
- Fiction: Anthropic's Pause Would Be the Most Expensive Alarm in Corporate History — DavidSKrueger · 2026-09-10