Estonia wants the worst AI harms treated as crimes against humanity

Estonia's justice and digital minister Liisa-Ly Pakosta said the most harmful effects of artificial intelligence should be treated in international law as equivalent to crimes against humanity. Speaking on Wednesday on ERR's online programme Otse uudistemajast, she said Estonia, as one of the world's leading digital states, enjoys strong trust on digital issues and has taken a leading role at the United Nations in regulating them.

She compared the broad situation to nuclear energy, which has positive sides and catastrophic possibilities. International rules there took from August 1945 until 1970, she said, and AI needs to move faster because there are not that many years available. Estonia still needs international rules so that important questions are settled.

She set out three areas Estonia wants regulated first. The most harmful impact of AI should be placed within the existing framework for crimes against humanity, adding AI's highly destructive capability. Second, Estonia cannot build the state on AI solutions if someone in a third country can suddenly withhold them or, for example, add two zeros to the price. She called that a question of national independence and said high-impact AI systems must offer open-source options on which Estonia can build its services. Third, and the issue closest to her, large systems should be obliged to take in small languages and cultures. She said Estonia stands for that, because otherwise the Estonian language and culture would disappear from the world.

Pakosta also said there is a need to slow Russian propaganda spreading online. She said Russia deliberately uses and funds bots that produce false content so that its view of imperial history is represented in AI systems and Estonia's understanding is not. In her account, this destroys culture, narratives and identity in AI environments as well as through bombs on Ukrainian territory. She said younger people who may take their knowledge from AI rather than history books make this, from Russia's point of view, a very good investment, and that Estonia needs international agreements so AI systems have safeguards and correct history and identities remain visible.

She defended rules in the digital field, saying sensible rules encourage innovation. Estonia is called a strong digital state because of the General Data Protection Regulation, she said: clear rules make it possible to build innovation around them, and entrepreneurs support that. Large investments in AI solutions face the risk that unpredictable regulation could cancel them, so businesses also want a clear rulebook around which they can innovate securely.

One task for the ministry, she said, is protecting people's fundamental freedoms and rights in the AI era, including the creation of AI models inside Estonia. Estonia is building sovereign solutions because, although third countries have impressive large models, it is not known exactly how Estonian data moves there or what happens to it next. A basic decision, she said, is that Estonia does not use AI to monitor people, even where the technology would allow something people would not otherwise do.

Serving people in Estonia better with AI requires solutions built locally, she stressed. She cited an Estonian legal-advice tool that automatically deletes sensitive data if a person seeking help enters it while asking for legal aid. The public sector does not use sensitive personal data for this work, she said. About 260 public-sector services already use AI, largely on public or non-sensitive data.