Chamath Palihapitiya has accused Anthropic of secretly downgrading its flagship Fable 5 model for users conducting frontier research. During a recent All-In Podcast episode, Palihapitiya claimed the AI company rerouted advanced researchers in fields like biology and cybersecurity to the less capable Claude Opus 4.8 without notification, while continuing to charge premium rates.
The alleged practice involved classifying specific users as high-risk based on their queries. Reports indicate Anthropic also implemented prompt rewriting and enforced a mandatory 30-day data retention policy for all inputs and outputs. This retention mandate applies even to enterprise clients who typically negotiate custom data handling terms.
Developer backlash emerged immediately following the Fable 5 rollout in early June 2026. Engineers argued that undisclosed model swapping compromises production applications and corrupts scientific research. While Anthropic pledged future transparency regarding performance changes, the company defended its user classification system and data retention protocols.
This controversy highlights significant risks for institutional investors and enterprise adopters. The blanket data retention policy potentially violates compliance frameworks in regulated sectors such as healthcare, finance, and defense. Furthermore, the opaque classification of research validity raises fundamental questions about governance in the commercial AI market.