Russia uses AI for drones, propaganda and surveillance — The Japan Times
In Russia — an aggressor state — artificial intelligence is increasingly being used for military, propaganda and surveillance purposes, although the country lags behind the United States and China in developing advanced models. The Japan Times reports.
AI is used in drones with which Russia attacks Ukraine, to create fake social media profiles and spread Kremlin narratives abroad, as well as in facial-recognition video surveillance networks inside the country. Publishing group Eksmo-AST, after strict censorship laws were introduced, employed AI to search books for passages that could conflict with the new restrictions. According to the group’s president, Oleg Novikov, the system conducted about 20,000 checks, while 600 publications were submitted to external experts for additional analysis.
Lagging behind in foundation models
Russia has at least two large language models: Alice by Yandex and GigaChat by Russia’s largest bank, Sberbank. Sberbank head German Gref said that the bank’s development lags Chinese models by about six months and the most advanced American ones by a year. He also claimed that GigaChat is the only foundation model created entirely by a Russian company and said that Yandex is fine-tuning the Chinese Qwen model. Yandex categorically denied this claim.
The publication’s sources, familiar with Russia’s AI capabilities, link the limitations to the war against Ukraine, sanctions, a shortage of high-performance accelerators, lack of access to non-Russian data sets, and the redirection of investment toward the war economy. According to them, Russia uses Chinese and Western open-weight models for certain military tasks.
More current news is available on the UA.News Telegram channel Telegram.
Tactical military developments
According to one of The Japan Times’ sources, who has direct knowledge of these efforts, Russia is working on AI-controlled drone swarms. Several Chinese open-weight models may be sufficient for such tasks because they do not require the most advanced artificial intelligence systems. In February, OpenAI blocked a group of ChatGPT accounts linked to the pro-Russian Telegram channel Rybar and its disinformation distribution network.
Russia’s Ministry of Defense, according to the publication, opened archives containing hundreds of terabytes of video from fiber-optic and radio-controlled strike drones to vetted domestic developers for AI model training. Kateryna Bondar, a senior fellow at CSIS, noted that Russia’s main limitation is not a lack of specialists or ambition, but computing power and components.
This year, the Kremlin also established a special commission for the development of key technologies, headed by presidential aide Maksim Oreshkin and Deputy Prime Minister Dmitry Grigorenko. On September 1, the first comprehensive AI law came into force, dividing approved models into “sovereign” and “national” ones. The latter may use foreign or open components but must be under Russian control and operate on Russian infrastructure.
