UK Freedom of Information Laws and AI Chatbots

The use of AI chatbots, including ChatGPT, by UK government departments and public bodies is now subject to Freedom of Information (FOI) laws. This shift allows for greater public access to records related to AI interactions, as confirmed by the Information Commissioner’s Office (ICO), the UK’s data protection authority.

Recent guidance from the ICO states that “when public sector employees utilize AI for work purposes, the information generated is subject to FOIA,” including the prompts used by these AI systems.

Significant Precedent Set

Last year, a breakthrough occurred when New Scientist successfully requested the ChatGPT logs from then UK technology secretary Peter Kyle under the FOI Act, marking a potential world first. This prompted various news organizations to seek further information, although many requests were either denied, citing cost reasons or a vague label of “It’s infuriating,” allowing authorities to refuse requests.

However, the clarification from the ICO could transform this landscape. As John Baines from Mishcon de Reya explains, “Currently, it would be very difficult for public authorities to argue that AI-related requests are not subject to FOIA.” He emphasizes that if information is “held” in any recorded form by a public authority, it is fundamentally subject to FOIA, including both inputs and outputs related to AI systems.

The Consensus Among Experts

“I think that should be uncontroversial,” states Tim Turner, a data protection expert based in Manchester, UK. “If records were collected by public officials performing their duties, they are within scope. This should apply equally to AI interactions and Post-it notes.”

The new guidelines may empower government employees to successfully request prompts used with AI tools. The ICO has also suggested that public authorities might be required to utilize AI to summarize large documents and datasets when responding to requests, potentially overcoming previous costs that led to request rejections.

Criticism and Concerns

Despite these advancements, some critics have voiced concerns regarding the use of FOI laws to access AI chat logs. Matt Clifford, chair of the UK Advanced Research and Inventions Agency (ARIA), expressed on social media that the decision to release Mr. Kyle’s interactions with ChatGPT is “absurd,” arguing that it may discourage ministers from using AI. Notably, ARIA enjoys exemptions from FOI laws.

When asked about the impetus for the new guidelines, New Scientist reported that the ICO did not provide a response. However, a spokesperson noted, “We regularly attend events and seek feedback on areas where both public authorities and requesters value further clarity and guidance. Our recent guidance on AI and FOI reflects what we’ve heard from organizations, and we’ve tested it with external stakeholders to ensure the content is as clear and useful as possible.”

Source: www.newscientist.com