top of page

Russian AI policy: The Trials and Tribulations of Digital Sovereignty

JAMIE EDWARDS | Fellow

Will the strategic importance of AI development save it from the Russian state’s need for control?

At the St Petersburg Economic Forum in June 2026, Russian president Vladimir Putin declared that the field of AI development in Russia “has very good prospects”.[1] Earlier this year, a draft bill regulating the use of AI in Russia was published for public discussion by the Ministry of Digital Development. Notably, this required AI to be trained only on Russian data and developed by Russian citizens.[2] This insistence on completely domestic production was a natural continuation of wider digital policy goals that aim to create an isolated and controlled digital ecosystem as an alternative to the existing global internet, using the framing of “sovereignty” to push for greater state control.[3.4]

 

However, the Ministry of Digital Development has now been forced to pull back on some of their more stringent requirements. After the initial draft bill was announced, Russian industry representatives and experts widely protested on the grounds that this would hinder technological development, arguing that a lack of Russian data and digital infrastructure made complying with the proposed legislation impossible.[5] Many existing Russian models are based on open-source models such as DeepSeek, and insistence on exclusively Russian-origin data and development would risk shutting these down and significantly raising development costs for firms.[6] 

 

As a result of these protests, the requirement that models with “sovereign” or “national” status must be only trained on Russian data and developed by Russian citizens has now been dropped from the revised draft bill.[7] This may appear at least a partial victory for Russian industry, but the state has proved perfectly willing to compromise on economic and technological development in favour of greater control since the beginning of the full-scale invasion of Ukraine. One striking example is the recurrent internet shutdowns that are suspected to be a result of the state drive to create a controlled digital space, even as these shutdowns cost local businesses billions of roubles.[8] While Russian AI development may be of such strategic importance that it is granted greater global access, it is also likely that while the most stringent demands have been removed from the current draft legislation, in the future state actors will find reasons to ban or restrict use of foreign datasets and models.

 

Furthermore, while this draft bill has been revised requirements to not only to comply with Russian law, but also to include “traditional Russian spiritual and moral values” are retained. This includes patriotism, service to the Fatherland and responsibility for its fate, historical memory, and unity of the Russian people, among others.[7] Perhaps unsurprisingly, when queried in Russian on facts related to the invasion of Ukraine, researchers found that Russian AI models not only reproduced lines from Russian propaganda, but also edited their own output to comply with these state narratives.[9]

 

Worryingly, even if the Russian state fails in its aims to create a 100% domestically produced AI, it is entirely possible that biases in its favour are already baked into existing AI models, even those developed outside the country. As shown in a recent paper in Nature, commercial models already favour state narratives when responding to queries in different languages, due to the preponderance of state media in training data. As a result, the authors note, “In intentionally shaping the media environment, powerful institutions may have unintentionally shaped the way LLMs generate text”.[10]

 

To what extent state demands for control will outweigh companies’ need to participate in global AI developments remains to be seen, but as it stands now, Russian AI policy is a compromise between the desire for a secure “walled garden” and the reality of global technological development. While the state appears to have given way on some of the stricter requirements, recent history shows its willingness to compromise development for the sake of maintaining domestic control.

Dr. Jamie Edwards

Fellow

Jamie recently completed a PhD in area studies at the University of Oxford. His dissertation looked at the socio-economic aspects of the Russian Army. At Oxford, he taught courses on politics and quantitative research methods. Jamie is interested in AI ethics, particularly how its integration into advanced academic research and everyday life raises questions of accountability, transparency, and impact.

Jamie also holds an MPhil in Russian and Eastern European Studies from the University of Oxford.

References

1. ‘Путин видит у России очень хорошие перспективы в области ИИ’, TASS, 5 June 2026. 

<https://tass.ru/politika/27669575>

2. Dariia Mykhailenko, ‘Inside Russia’s “Sovereign AI” Plan—And Why It May Not Work’, UNITED24 Media, 17 April 2026.

<https://united24media.com/latest-news/inside-russias-sovereign-ai-plan-and-why-it-may-not-work-18002>

3. Maria Lipman, ‘Disconnecting the Russian Internet: Implications of the New “Digital Sovereignty” Bill’, Commentary, PONARS Eurasia, 21 February 2019.

<https://www.ponarseurasia.org/disconnecting-the-russian-internet-implications-of-the-new-digital-sovereignty-bill/>

4. Ivan U. K. Klyszcz, ‘The Double Edge of Russian Sovereign AI: Isolation and Narrative Consistency’, International Centre for Defence and Security, 11 May 2026.

<https://icds.ee/en/the-double-edge-of-russian-sovereign-ai-isolation-and-narrative-consistency/>

5. В «Роснефти» заявили о невозможности создать по требованию Путина «суверенный» ИИ’, The Moscow Times, 17 April 2026.

<https://ru.themoscowtimes.com/2026/04/17/v-rosnefti-zayavili-o-nevozmozhnosti-sozdat-po-trebovaniyu-putina-suverennii-ii-a192989>

6. Valery Kodachigov, ‘Отборная Модель: В России Не Будут Запрещать Зарубежные Нейросети’, Известия, 22 June 2026.

<https://iz.ru/2119376/valerii-kodachigov/v-rf-ne-budut-zapreshchat-zarubezhnye-neiroseti>

7. Власти скорректировали законопроект об ИИ после критики бизнеса’, RBK, 21 June 2026.

<https://www.rbc.ru/technology_and_media/22/06/2026/6a37ddb19a7947c93443fdc4>

8. Natalya Kovaleva, ‘Moscow Internet Blackouts: The Kremlin Tightens Its Grip on Russia’s Digital Space’, Chatham House, 27 March 2026.

<https://www.chathamhouse.org/2026/03/moscow-internet-blackouts-kremlin-tightens-its-grip-russias-digital-space>

9. Ihor Samokhodskyi, ‘I Tested Russia’s AI. It Knows the Truth, but It’s Been Trained to Lie’, The Kyiv Independent, 5 February 2026.

<https://kyivindependent.com/how-russia-turned-ai-into-a-cognitive-weapon-or/>

10. Hannah Waight et al., ‘State Media Control Influences Large Language Models’, Nature, ahead of print, 13 May 2026.

<https://doi.org/10.1038/s41586-026-10506-7>

bottom of page