Anthropic Accused of Secretly Downgrading Models, Sparking AI Transparency Debate
sull · x · 2026-08-13
Users have accused Anthropic of secretly downgrading services after the release of Fable 5 while still charging full price, alongside a policy of storing all prompts for 30 days. The developer community is reacting strongly, calling it the biggest violation of trust in AI history.
This has sparked a broader discussion about AI companies operating as a "Blackbox-as-a-Service." Some suggest it's entirely plausible technically for AI providers to swap premium models for cheaper ones during peak hours to save resources without transparency.
More from Companies & People
- Ramp AI Index Shows Anthropic Pulling Further Ahead of OpenAI — xeophon · 2026-08-13
- Non-Technical Solo Founder Raises $100M+ to Build AI Accounting Software — julianweisser · 2026-08-13
- AI safety org METR calls for talent, urging frontier AI employees to join — ajeya_cotra · 2026-08-13
- Pichai Counters AI Wall Theory: Massive Headroom Across Pre-training to Agentic Capabilities — r0ck3t23 · 2026-08-13
- ICONIQ report: Hybrid GTM becomes default, AI drives efficiency in 2026 — MartinGTobias · 2026-08-13
- YC CEO Garry Tan: AI Rewrites Startup Economics, Small Teams to Rival Giants via Agents — a16z · 2026-08-13