Russian IT Companies Ask Regulator to Allow Using Anonymized Data for AI Training Without User Consent

The Big Data Association, uniting companies such as Avito, Sber, Rostelecom, Yandex, and VK, has sent a letter to Russia's Ministry of Digital Development requesting amendments to the personal data law, as reported by Kommersant. The companies want permission to process data using privacy-enhancing technologies (PETs) without citizen consent for the development, training, and use of artificial intelligence. Currently, anonymized data is legally considered personal data, which the market players believe slows down AI development in Russia. The association argues that with correct application, identifying a specific person becomes impossible, but PETs are not recognized as a separate category in current legislation, as any algorithmic transformation of personal data is classified as 'anonymization methods.' The association proposes allowing PETs without consent under conditions: the risk of de-anonymization is zero or minimal, data is used exclusively for AI tasks, and processing occurs within Russia by Russian companies. Additionally, they suggest introducing a 'trusted intermediary' – a state-accredited structure that would store and provide access to anonymized data for AI needs, and propose separating data into paid (for commercial use) and free (for legal compliance). Svetlana Zakharova, business development director at Just AI, noted that companies have so far used data from internal sources, subsidiaries, and partner projects but this is no longer sufficient. A representative from MWS AI (part of MTS Web Services) said that expanding data access would accelerate the development of applied AI products. Vladimir Arlazarov, CEO of Smart Engines, explained that what is valuable for AI training is not identifying information but anonymized datasets, such as behavioral and medical statistics, while some needs are already covered by synthetic data generation. In July 2026, the State Duma adopted a law on AI development in Russia, which for the first time establishes a legal framework for developing large models and labeling AI content, including allowing training neural networks on state data and works without separate licensing.