You can FOI (Freedom of Information) ChatGPT now

This week the government has – possibly inadvertently – set a precedent regarding how AI chatbot prompts will be treated under the Freedom of Information (FOI) Act. In a first-of-its-kind case, the New Scientist successfully obtained chatbot interactions from Science, Innovation and Technology Secretary Peter Kyle, under FOI laws, raising questions about whether AI-generated content should be subject to the same scrutiny as emails and messages or treated like private search engine queries. 

Kyle’s prompts to ChatGPT, as reported by the New Scientist, included requests for insights on why UK businesses have been slow to adopt AI, recommendations for podcasts where he should make appearances, and definitions of technical terms relevant to his role. The chatbot provided structured responses, highlighting key barriers to AI adoption such as regulatory concerns, lack of awareness, and insufficient government support. 

Initially, the Department for Science, Innovation and Technology (DSIT) resisted the FOI request, arguing that Kyle’s interactions included both personal and official usage. However, upon refinement of the request to focus solely on prompts made in his official capacity, DSIT complied—an outcome which appears to have surprised both the publication and data protection experts, who anticipated government reluctance to establish this sort of precedent. 

The release of these records has sparked debate about how AI chatbot interactions should be classified under FOI laws. Historically, emails and WhatsApp messages have been deemed eligible for disclosure, while search engine queries have not. The key question is whether AI chatbot prompts align more closely with one category or the other. 

By complying with the New Scientist’s FOI request, DSIT has set a significant precedent that could shape future requests for chatbot interactions – and impact how government figures interact with AI chatbots like ChatGPT. The government might have been wiser to make more of an effort to argue that AI-generated responses fall outside the scope of FOI, as this decision is likely open the door to further disclosures. Given the Prime Minister’s stated interest in integrating AI into government decision-making, this case could ultimately prove to be a significant oversight.