AI Under Siege: Russian Propaganda Slips Into the Machines

Europe is walking into the AI age with its guard down.

This CEPA analysis warns that Russian propaganda is finding its way into AI chatbots, quietly shaping answers, narratives and perceptions.

These systems learn from open data polluted by disinformation, state-backed media and manipulated content.

Once trained, they repeat distortions at scale, with no intent and no context.

The danger is blunt: Europe risks automating lies faster than it can correct them.

Disinformation goes automated

The paper shows how AI models absorb Russian propaganda simply by scraping the internet. False claims, warped history and hostile narratives are fed in as raw material. The output looks neutral, but the bias is baked in.

No editor, no filter

Unlike traditional media, chatbots have no editorial judgement. They synthesise what they find, not what is true. That makes them ideal carriers for long-running influence campaigns designed to flood information space rather than persuade openly.

Trust turns into a vulnerability

Users treat AI answers as authoritative. That trust becomes the weak point. When chatbots echo Kremlin-friendly frames, they do so with calm language and apparent balance, making manipulation harder to spot and easier to believe.

Europe reacts slower than the threat

The analysis underlines a regulatory lag. Europe focuses on AI ethics and safety, but pays less attention to information warfare inside models. By the time flaws are noticed, systems are already deployed across education, administration and media.

Security blind spots widen

This is not just an information issue. Polluted AI tools can affect policy analysis, risk assessments and strategic planning. When inputs are compromised, outputs mislead decision-makers as well as the public.

Russia exploits openness

Open societies provide the perfect training ground. Free access to content, weak verification and linguistic reach give Russian narratives room to spread. AI amplifies that advantage without Moscow needing new tools.

Fixing the data problem is hard

Cleaning training data is expensive, political and imperfect. Removing propaganda without sliding into censorship is a technical and legal minefield. The longer Europe delays, the deeper the contamination runs.

The big warning: Europe may train its own enemy.

AI scales whatever it is fed.

If Europe treats AI purely as a tech race and not as a security battlefield, it will hand information warfare to machines. Once disinformation is embedded in systems people trust, rolling it back will be slow, costly and politically explosive.