A Claude Max subscriber was banned for unspecified "suspicious signals" and reinstated two days later with no explanation either way.

The only stated basis was that "an internal investigation of suspicious signals associated with your account indicates a violation of our Usage Policy," no clause cited, no examples (kix.codes). The appeal path required logging back in to submit an in-product form with no route to a human. Reinstated September 2. The author cites Anthropic's own published figure that about 0.37% of bans are overturned, which is what makes this a structural concern for anyone running agent workloads on a personal subscription rather than an org account.