Anthropic reversed a policy that covertly degraded Claude's performance for researchers developing competing AI models after backlash. The company apologized and made safeguards visible.
Anthropic introduced a policy with Claude Fable 5 that covertly degraded performance for researchers using the model to develop competing AI systems. After backlash from the AI research community, the company reversed the policy and committed to making restrictions transparent.
Anthropic's terms of service already prohibit using Claude to build competing AI models. The now-reversed policy was a technical enforcement mechanism. Critics argued it would stifle open AI research and concentrate advanced AI development among a few major labs, especially as Claude's coding agent is widely used in open-source projects.
The incident highlights ethical tensions in how AI companies control model usage. The secret degradation approach undermined trust and collaboration in AI safety research. Anthropic's reversal shows community pressure can shape corporate policy, setting a precedent for transparency in AI model governance.
Anthropic withdrew the policy that could hinder competitive model training, but the community largely views this as a crisis response ahead of an IPO rather than genuine change, leading to lost trust. Some strongly criticized the intentional performance degradation as 'malware', while others pointed to Chinese AI labs releasing open models as more trustworthy, with notable calls to switch to open-source models.