Sārts: Russia Is Poisoning Large Language Models in Context of Latvia's Elections

Jānis Sārts, director of NATO's Strategic Communications Centre of Excellence (StratCom COE), addressed the issue of Russian informational influence on Latvia in an interview with the magazine "Ir." He said Russia is deliberately acting, in the context of Latvia's elections, to influence — or, in his words, "poison" — large language models, the artificial intelligence systems increasingly used by people to search for information and ask questions in daily life.

About StratCom COE

Sārts heads the NATO body headquartered in Riga that studies strategic communications and disinformation, helping alliance member states recognize and resist foreign informational influence. His warning suggests NATO experts see Russian influence operations extending beyond traditional media and social networks into newer technologies that the public increasingly relies on.

A newer tool in influence operations

The claim made in the "Ir" interview indicates that manipulating artificial intelligence tools, including language models, is regarded as one of the newest instruments Russia is using to shape public opinion and political processes in Latvia, including around elections.

The publicly available excerpt does not provide further detail on the specific mechanisms involved or which particular election Sārts referenced in the interview. However, the warning adds to previously voiced concerns from NATO and Latvian security institutions about Russian hybrid threats and information warfare directed at the Baltic states.