Is AI a threat to privacy?
AI can leak personal data, infer sensitive information and make surveillance such as facial recognition cheaper. Widespread privacy breaches from chatbots have not yet been documented, but misuse of facial recognition has, including by Swedish police.
The main privacy risks
The International AI Safety Report 2025 describes three ways general-purpose AI can harm privacy. Sensitive information in the training data can leak when someone uses the system. Information users type in can also leak. And AI can be used deliberately to piece together sensitive facts about specific people from large amounts of data.
The same report notes that researchers have so far not found evidence of widespread privacy violations linked to general-purpose AI. The concern is therefore partly about what could happen as the systems spread into more services that handle personal data.
Facial recognition: a Swedish case
Surveillance is where the risk is most concrete. In February 2021 the Swedish Data Protection Authority (IMY) found that employees of the Swedish Police Authority had used Clearview AI, a facial recognition service that matches photos against a very large number of images scraped from the internet, from autumn 2019 to March 2020 without a legal basis. IMY fined the police SEK 2.5 million and ordered them to train staff, inform the people whose images had been uploaded where possible, and try to get the data deleted.
What is being done
The GDPR applies to AI just as to other data processing. In December 2024 the European Data Protection Board said that an AI model trained on personal data is not automatically anonymous; each case has to be assessed. If a model was built with unlawfully processed personal data, that can affect whether it may lawfully be used.
The EU AI Act has, since February 2025, banned several surveillance practices: building facial recognition databases by untargeted scraping of the internet or CCTV footage, emotion recognition in workplaces and schools, biometric categorisation to infer sensitive characteristics, and in principle real-time remote biometric identification by police in public spaces.
What you can do
- Avoid typing sensitive personal information about yourself or others into AI chatbots, especially at work.
- Check the privacy settings of the AI services you use, including whether your conversations may be used for training.
- Use your GDPR right of access to ask a company what personal data it holds about you and how it is used.
- Think about which photos and voice recordings of you are publicly available online.
Sources
- International AI Safety Report 2025 – International AI Safety Report (chair: Yoshua Bengio), 2025
- Beslut efter tillsyn enligt brottsdatalagen – Polismyndighetens användning av Clearview AI (DI-2020-2719) – Integritetsskyddsmyndigheten (IMY), 2021
- EDPB opinion on AI models: GDPR principles support responsible AI – European Data Protection Board (EDPB), 2024
- AI Act – regulatory framework for AI – European Commission, 2026
Data updated: 2026-11-19