Swedish Prime Minister Ulf Kristersson has stirred up public debate over politicians’ use of artificial intelligence (AI) after telling local media he uses ChatGPT to brainstorm and seek a “second opinion” on how to run the country.
Kristersson told the Swedish newspaper Dagens Industri that he uses ChatGPT and the French service LeChat, and that his colleagues also use AI in their everyday work.
“I use it myself quite often, if for nothing else than for a second opinion. ‘What have others done? And should we think the complete opposite?’ Those types of questions,” he said.
The comment sparked backlash, with critics arguing that voters had elected Kristersson, not ChatGPT, to lead Sweden.
Technology experts in Sweden have since raised concerns about politicians using AI tools in such a way, citing the risk of making political decisions based on inaccurate information.
Large language models’ (LLMs) training data can be incomplete or biased, causing chatbots to give incorrect answers or so-called “hallucinations”.
“Getting answers from LLMs is cheap, but reliability is the biggest bottleneck,” Yarin Gal, an associate professor of machine learning at the University of Oxford, previously told Euronews Next.
Experts were also concerned about sensitive state information being used to train later models of ChatGPT, which is made by OpenAI. Its servers are based in the United States.
Kristersson’s press team brushed aside security concerns.
“Of course, it's not security-sensitive information that ends up there. It's used more as a sounding board,” Tom Samuelsson, Kristersson’s press secretary, told the newspaper Aftonbladet.
Should politicians use AI chatbots?
This is not the first time a politician has been placed under fire due to their use of AI – or even the first time in Sweden. Last year, Olle Thorell, a Social Democrat in Sweden’s parliament, used ChatGPT to write 180 written questions to the country's ministers.
He faced criticism of overburdening ministers’ staff, as they are required to answer within a set time frame.
Earlier this year, United Kingdom tech secretary Peter Kyle’s use of ChatGPT came under fire after the British magazine, New Scientist revealed he had asked the chatbot why AI adoption is so slow in the UK business community and which podcasts he should appear on to “reach a wide audience that’s appropriate for ministerial responsibilities”.
Some politicians make no secret of their AI use. In a newspaper column, Scottish Member of Parliament Graham Leadbitter said he uses AI to write speeches because it helps him sift through dense reading and gives him “a good basis to work from” – but emphasised that he still calls the shots.
“I choose the subject matter, I choose the evidence I want it to access, I ask for a specific type of document, and I check what’s coming out accords with what I want to achieve,” Leadbitter wrote in The National.
And in 2024, the European Commission rolled out its own generative AI tool, called GPT@EC, to help staff draft and summarise documents on an experimental basis.
ChatGPT available to US public servants
Meanwhile, OpenAI announced a partnership this week with the US government to grant the country’s entire federal workforce access to ChatGPT Enterprise at the nominal cost of $1 for the next year.
The announcement came shortly after the Trump administration launched its AI Action Plan, which aims to expand AI use across the federal government to boost efficiency and slash time spent on paperwork, among other initiatives.
In a statement, OpenAI said the programme would involve “strong guardrails, high transparency, and deep respect” for the “public mission” of federal government workers.
The company said it has seen the benefits of using AI in the public sector through its pilot programme in Pennsylvania, where public servants reportedly saved an average of about 95 minutes per day on routine tasks using ChatGPT.
“Whether managing complex budgets, analysing threats to national security, or handling day-to-day operations of public offices, all public servants deserve access to the best technology available,” OpenAI said.
