Anthropic blocks users for anything touching training larger models

A user identified as Jiahao Shao publicly accused Anthropic of blocking accounts and downgrading model access for activity it tagged as "training a competitor," writing on X that "anything touching RFT, SFT, or training larger models gets me blocked, Opus 5.5 downgraded to Opus 5, or the session killed outright." Shao denied distilling Claude, stating the rollouts came from a strong open-source model and that Claude only helped submit jobs and run analysis, and criticized Anthropic CEO Dario Amodei for training on open web data while restricting paying users who do research automation. The complaint centers on Anthropic's enforcement of its anti-distillation policy against a paying researcher, which Shao characterized as a double standard protecting a commercial moat.

Jiahao Shao on X: "Shame on @DarioAmodei / Anthropic. I never distilled Claude—rollouts came from a strong open-source model; Claude just helped submit jobs and run analysis. Still got tagged for "training a competitor," and now anything touching RFT, SFT, or training larger models gets me blocked,… / X Jiahao Shao on X: "Shame on @DarioAmodei / Anthropic. I never distilled Claude—rollouts came from a strong open-source model; Claude just helped submit jobs and run analysis. Still got tagged for "training a competitor," and now anything touching RFT, SFT, or training larger models gets me blocked, Opus 5.5 downgraded to Opus 5, or the session killed outright. You built your models on the world's data, then treat paying users doing research automation as the enemy. Pathetic." Shame on @DarioAmodei / Anthropic. I never distilled Claude—rollouts came from a strong open-source model; Claude just helped submit jobs and run analysis. Still got tagged for "training a competitor," and now anything touching RFT, SFT, or training larger models gets me blocked, Opus 5.5 downgraded to Opus 5, or the session killed outright. You built your models on the world's data, then treat paying users doing research automation as the enemy. Pathetic. Shame on @DarioAmodei / Anthropic. I never distilled Claude—rollouts came from a strong open-source model; Claude just helped submit jobs and run analysis. Still got tagged for "training a competitor," and now anything touching RFT, SFT, or training larger models gets me blocked, Opus 5.5 downgraded to Opus 5, or the session killed outright. You built your models on the world's data, then treat paying users doing research automation as the enemy. Pathetic. And the anti-distillation crusade is peak double standard. You train on the open web—books, code, forums, everyone else's work—then ban anyone from learning from your outputs at scale. Protecting a moat is one thing; moralizing it while standing on scraped knowledge is another. I get why small startups refuse buyouts even when big tech waves cash—they'd rather stay independent than inherit that corporate value system. Anthropic sells itself as that independent alternative, then runs the same rigid enterprise lockdown on paying researchers. ClaimingShow more anthropic trained on the open web then blocked the user who tried to train on the outputs. the moat is the part that still needs the paying researcher to stay outside it. 🤡