GitHub announced that, starting from April 24, 2026, it will update its codebase policy to utilize interaction data from Copilot Free, Pro, and Pro+ users to train AI models. This data includes information such as model inputs and outputs. The Chief Product Officer stated that this move aims to enhance the accuracy and security of code suggestions, with internal Microsoft tests showing an increase in the acceptance rate of suggestions. The policy adopts an 'opt-out' mechanism, sparking discussions within the developer community regarding the definition of private repositories and data ownership rights. However, Copilot Business, Enterprise users, and educational users bound by contractual terms are not affected by this policy update. GitHub claims that this move aligns with industry practices, but incorporating code from private repositories into the training set undoubtedly poses a challenge to the concept of 'privacy.' This policy shift signifies that GitHub is gradually moving towards a closed-loop AI training ecosystem and also heralds a new phase in the field of AI developer tools, where data compliance and model evolution will engage in mutual game theory (Note: ' game theory ' is translated as 'mutual interaction' or 'competition' in context, but kept as ' game theory ' in pinyin for precision; a more natural English equivalent might be 'where data compliance and model evolution will be in a dynamic interplay'). (For ' game theory ', a more contextually fitting translation could be 'where data compliance and model evolution will engage in a delicate balance or competition') A more refined sentence could be: '...heralds a new phase where data compliance and model evolution will be in a constant interplay.'
