Andorran Data Agency Urges Caution with AI Chatbots to Protect Personal Data

The Andorran Data Protection Agency has urged residents to approach artificial intelligence tools, particularly chatbots, with caution to safeguard personal data amid their growing popularity.

In an opinion piece published by *Diari d'Andorra*, Anna Coma, the agency's data protection inspector, draws parallels between historical resistance to calculators in the 1980s and current hesitations over AI. She references a 1986 *Washington Post* article detailing maths teachers' protests against integrating calculators into school curricula for classwork, homework, and assessments. Coma argues that, just as calculators freed up time for deeper reflection by simplifying computations, AI chatbots—powered by large language models—offer similar potential but demand careful handling.

These models operate as probabilistic systems trained on vast datasets, including personal information, to predict word sequences rather than verify facts. This limitation can lead to "hallucinations," or inaccurate outputs, Coma explains. Crucially, she emphasises that data protection must be embedded from the outset, both by developers and users.

Key recommendations include configuring AI tools with privacy by design and default settings: implementing robust security measures, providing transparent information on data processing, and securing explicit user consent. For individuals, Coma advises reviewing privacy settings on devices and services, enabling two-factor authentication, and actively protecting personal data.

The agency stresses that compliance by Andorran businesses and entities, combined with public awareness, will help mitigate risks in an increasingly digital world. Coma concludes that mastering these tools securely, much like past technologies, requires overcoming fear of the unknown through informed use.