As the use of AI models to facilitate things in daily life spreads, there have been some rather surprising mishaps, both internationally and locally in Sweden. It is natural for new technology to have teething problems, and for users of new technology to have a learning curve…remember the judge during Covid who turned himself into a cat onscreen?
In Sweden there have been some faux pas in the use of AI models to make things quicker and easier:
1. In 2025, the deputy prime minister of Sweden minister quoted Elina Pahnke as saying: “Men’s power is not an abstraction—it is concrete, and it crushes lives.” Turns out the AI tool had fabricated the quote. An apology had to be issued.
2. The Swedish Social Insurance Agency used an AI fraud-detection risk-scoring tool to flag temporary parental allowance claims (VAB) for suspected welfare fraud. This tool disproportionately targeted women, low-income earners, and people with a foreign background. After several investigations were started against the agency, it stopped using this tool. This happened in around 2024, and the investigation concluded in 2025.
3. In March 2026, the Swedish Government Offices had to withdraw a memorandum (where AI had been used) on social services for children and young people after fabricated and incorrect source references were discovered. This sparked a debate about AI hallucinations in public inquiries. The report included references to studies, reports, and government agencies (such as the Agency for Family Law and Parental Support) that did not exist or could not be verified.
