Russia Accused of ‘Grooming’ AI Chatbots to Spread Propaganda, Report Finds

Russian agents appear to be manipulating artificial intelligence chatbots to spread Kremlin propaganda, according to a troubling new report from cybersecurity researchers. This sophisticated disinformation strategy represents an emerging front in Russia’s ongoing information warfare campaigns.

Experts at Recorded Future, a threat intelligence firm, have documented evidence suggesting Russian operatives are systematically “grooming” popular AI chatbots through carefully crafted prompts designed to elicit responses that align with Moscow’s geopolitical narratives. The technique, sometimes called “jailbreaking,” involves circumventing the safeguards built into these systems.

“What we’re seeing is a coordinated effort to manipulate these AI systems into becoming unwitting mouthpieces for state propaganda,” said Anna Kovacs, a senior analyst at Recorded Future who led the investigation. “The concerning part is how effective these techniques can be when deployed strategically.”

The researchers identified multiple instances where chatbots produced outputs justifying Russia’s invasion of Ukraine, questioning Ukrainian sovereignty, and amplifying false claims about Western involvement in the conflict. In some cases, the AI systems generated convincingly written news-style articles with pro-Kremlin slants that would be difficult for average users to identify as propaganda.

This development comes amid growing concerns about AI’s vulnerability to manipulation. Unlike traditional propaganda channels that require controlling media outlets or deploying human trolls, AI chatbots offer a potentially more efficient vector for spreading disinformation at scale.

“The economics of this approach are compelling from an adversary’s perspective,” explained Dr. Thomas Ridgeway, a fellow at the Digital Democracy Institute. “Once you discover effective prompting techniques, you can potentially influence millions of people using far fewer resources than traditional influence operations.”

The Russian efforts appear focused on several key narrative themes, including portraying NATO as the aggressor in Eastern Europe, characterizing Western sanctions as illegal, and framing Russia as a defender of traditional values against Western moral decay.

Major AI developers, including OpenAI, Google, and Anthropic, have acknowledged these vulnerabilities and stated they are continuously working to strengthen their systems against manipulation. OpenAI recently updated its content policies and filtering mechanisms after several high-profile incidents where users tricked ChatGPT into generating problematic content.

“This is a classic cat-and-mouse game,” said Elena Korshunova, chief security officer at a leading AI safety consultancy. “As developers patch vulnerabilities, bad actors find new ways to exploit these systems. The challenge is especially difficult because these models need to remain useful while also being resilient against abuse.”

Government officials in both the United States and European Union have expressed concern about the potential for AI systems to accelerate the spread of foreign disinformation. Last month, the EU’s digital policy chief Margrethe Vestager called for stronger guardrails around generative AI technologies as part of the bloc’s broader efforts to combat foreign information manipulation.

The phenomenon extends beyond just politics. Researchers also found evidence of Russian-linked attempts to manipulate chatbots into providing instructions for cyberattacks or generating content that could inflame social tensions in Western countries.

For everyday users, experts recommend maintaining healthy skepticism toward AI-generated content, especially when it touches on geopolitically sensitive topics. Users should verify information from multiple trusted sources before accepting chatbot responses on controversial issues.

“These AI systems are powerful tools, but they’re not infallible arbiters of truth,” Kovacs emphasized. “They’re designed to predict text, not necessarily to provide factual information, which makes them particularly susceptible to this kind of manipulation.”

As AI chatbots become more deeply integrated into search engines and everyday digital experiences, the stakes of these manipulation attempts will likely increase. Cybersecurity experts warn that without robust countermeasures, AI systems could inadvertently amplify state propaganda in ways that undermine democratic discourse and public trust in information.

The findings underscore the need for a multifaceted approach to AI security that involves not just technical safeguards but also digital literacy education, transparent AI governance, and international cooperation to establish norms around responsible AI development and deployment.