Anthropic Faces Developer Backlash Over Model Training Restrictions and Account Bans

Researchers report account restrictions and model downgrades for workflows touching reinforcement fine-tuning or training larger models.

Developers leveraging Claude for research automation have voiced strong frustration following automated account restrictions and model tier downgrades. Users report that workflows interacting with reinforcement fine-tuning (RFT), supervised fine-tuning (SFT), or broader model training activities have triggered anti-competition flags.

Critics argue that penalizing paying researchers who use API capabilities for legitimate analysis stifles innovation, highlighting ongoing industry friction around model distillation policies and platform guardrails.