Estonia wants the worst AI harms treated as crimes against humanity

Justice and Digital Affairs Minister Liisa Pakosta has called for stronger international regulation of artificial intelligence to protect national sovereignty, smaller languages and cultures. Speaking on ERR's "Otse uudistemajast" webcast, she compared AI with nuclear energy: both can bring benefits and both can cause catastrophes. She said international nuclear rules took from August 1945 until 1970 to establish, and that AI cannot wait that long.

Pakosta outlined three Estonian priorities. First, the most harmful impacts of AI should be treated as crimes against humanity, extending an existing legal framework to AI's extremely harmful and destructive capabilities. Second, Estonia's digital infrastructure must not depend on technology controlled by foreign providers. She said the state cannot be built around AI solutions that a third country could withdraw or sharply reprice, and that systems with a significant impact need open-source options Estonia can build services on. Third, large AI systems should be required to include small languages and cultures, which she said Estonia is fighting for so that the Estonian language and culture do not disappear.

She warned that Russia is deliberately financing bots so that its interpretation of imperialist history enters AI systems while Estonia's perspective does not. She said Russia is destroying cultures with bombs on Ukrainian territory and is also erasing culture, narratives, stories and identity in AI environments. Younger generations increasingly rely on AI rather than history books, she said, which could make such manipulation an effective investment. She called for international agreements requiring safeguards against disinformation and distorted historical narratives.

Pakosta argued that clear rules encourage innovation and that businesses support them, pointing to Estonia's experience with the EU General Data Protection Regulation. Companies investing heavily in AI need predictable rules to protect those investments, she said.

Another priority is sovereign AI so residents' personal information is protected. She said impressive large models exist in third countries, but Estonia does not know exactly how its data moves through those systems or what happens to it afterward. Estonia has decided not to use AI for surveillance that technology makes possible but that people would not otherwise accept. The country currently uses AI in about 260 public-sector services, relying on public or non-sensitive information. As a local example, she cited an Estonian legal advice service that automatically deletes sensitive personal information entered by users.