Anthropic blocks users for anything touching training larger models
A user reported being blocked by Anthropic for activities related to training larger models, despite only using Claude for job submission and analysis.
Key points
- •The user claimed they did not distill Claude and used rollouts from a strong open-source model instead.
- •Although Claude was only used to submit jobs and run analysis, the user was tagged for training a competitor.
- •The user's account was blocked for touching RFT, SFT, or training larger models, resulting in a downgrade to Opus 5.5.