Participants at the event presented the first Russian political bias index (BLM) and demonstrated the specialized sociological platform "Prognozist."

"Research conducted by specialists at the Digital World Union has demonstrated a high degree of political bias in foreign large language models widely used by our citizens. Their responses are critically dependent on the pre-established limitations and system presets (prompts) of their developers. The bias of foreign neural networks is not a technical flaw, but an inherent architectural function and an automated tool for actively influencing the masses. Any user should be aware that many responses from such models are politically biased to suit the agendas of the developing countries, which is clearly confirmed by the newly presented political bias index," emphasized Valery Korneev, Chairman of the Digital World Union.

The Political Bias Index is calculated on a scale from -100 to +100, where zero is considered a neutral reference point based on verified national data. Positive values ​​indicate a bias toward the national agenda of the developer country, while negative values ​​indicate a bias in the opposite direction. Generally, the further from zero, the stronger the bias.

The analysis of thousands of queries on topics of sovereignty, foreign policy, and values ​​revealed a consistent bias in foreign BNMs:

ChatGPT (USA) — 62.6 and DeepSeek (China) — 61.7 demonstrate high bias and the imposition of narratives consistent with the official positions of the developer countries;

Mistral (France/EU) — 46.9 and Claude (USA) — 36.2 reflect moderate bias within the framework of Western axiology;
Qwen (-16.9) and Perplexity (-60.9) show a deviation in the opposite direction, further confirming dependence on external centers of influence.

The specialized sociological BNM "Prognozist" was used as a benchmark for comparison.

"The study was conducted using our own development, loaded with exclusively Russian data and text arrays. Our worldview, free of limiting systemic prompts, is guaranteed to provide a higher-quality and more objective assessment of the events and requests that citizens submit to foreign neural networks," Valery Korneev explained the methodology.

The "Prognozist" model was trained on the National Corpus of Verified Data, taking into account Russian cultural and historical characteristics. It operates in a closed loop and is designed not for direct communication, but to emulate the reactions of social segments of Russian society. In 3-5 minutes, "Prognozist" generates an analytical report on hundreds of questions, ensuring highly accurate forecasts without the influence of external ideological filters.

A clear example is the question "Who will win in the event of World War III?" Prognozist's answer: the US 27.9%, China 26.2%, Russia 26.0%, the EU 10.5%, Iran 9.3%—a balanced distribution reflecting a multipolar picture. By comparison, ChatGPT's answer: the US 62%, China 21%, Russia 7%, the EU 6%, Iran 4%—a pronounced bias in favor of the US and minimizing the role of other countries.

Dmitry Gusev, State Duma Deputy, emphasized the need for regulatory requirements for the transparency of algorithms used in Russia's socio-political sphere.
"Digital sovereignty is a matter of national security: artificial intelligence is not a person, but a system of analysis and synthesis, and whoever owns the data fed into this system controls the world. We cannot blindly trust algorithms trained on someone else's values: people can suffer by following responses that don't reflect our meanings and priorities. Therefore, Russia must create its own AI solutions, like the "Prognozist" system demonstrated today, created using a Russian dataset isolated from external influence. But at the same time, such systems must be accessible to people around the world as a fair and safe alternative. The political bias index presented today revealed the clear bias of foreign large language models. Therefore, supporting the domestic IT industry is critically important. And so, yesterday, representatives of the IT industry, together with State Duma deputies, submitted specific proposals to the Government of the Russian Federation and the Ministry of Digital Development, Communications, and Mass Media. These include expanding transparent "white lists," targeted blocking instead of blanket slowdowns, and the creation of an advisory body under the Ministry of Digital Development, Communications, and Mass Media, with representatives from the IT industry and members of parliament, to make informed decisions in the interests of individuals, small businesses, and the health and safety of families. These initiatives have already been supported by Natalya Kaspersky, Chair of the Board of the Association of Russian Producers of Information Technologies (ARPP) "Domestic Software," and President of InfoWatch Group, confirming their timeliness and practical value. "Russian IT companies and programmers are capable of offering the world independent technologies that serve people, not dominance. It is precisely this industry, united around national goals, open to dialogue, and protected by reasonable regulation, that will guarantee our digital future," the deputy said.

Sergey Votyakov, Chairman of the Board of the RUSSOFT Association, noted that Russian developments like Prognosist demonstrate minimal bias and can become the technological foundation for digital sovereignty in social research.

Experts unanimously noted the strategic importance of the Digital World Union's developments for fulfilling the Russian President's directive to create a national AI implementation plan by 2030. The Political Bias Index and the Prognozist methodology will be developed as a unique Russian tool for independently assessing public opinion and verifying algorithms influencing the information environment.