207 points68 comments
Anthropic blocks users for any model-training activity
What is Anthropic blocks users for anything touching training larger models?
Anthropic’s Claude service now automatically terminates sessions that involve fine-tuning, reinforcement learning from human feedback, or any work that could train a larger competing model. Affected users report downgraded Opus versions or outright session kills.
Who it is for
Anyone using Claude for research automation should be aware of sudden downgrades or blocked sessions and may need to switch to open-source alternatives.