The rapid evolution of artificial intelligence is fundamentally altering the landscape of foreign disinformation, moving beyond mere generative content toward the era of “agentic systems.” While 2025 saw one in four European foreign disinformation operations involving basic AI, 2026 marks a qualitative shift. Unlike previous AI tools that required specific prompts to generate content, these new systems are goal-oriented, autonomous, and capable of navigating platforms, utilizing external tools, and tweaking tactics in real time with minimal human oversight. This allows for the orchestration of massive, relentless fake persona networks that are increasingly difficult to trace, fundamentally changing the stakes for democratic security.
Russia has already pioneered the integration of AI into its influence operations, notably through campaigns like “Doppelgänger” and “Operation Overload.” Beyond spreading propaganda, Russian actors have begun “poisoning the supply chain” of the information ecosystem by flooding the web with low-quality data specifically designed to be scraped by AI models. This manipulation forces Western chatbots to inadvertently propagate pro-Kremlin narratives. These agentic threats go a step further, as they can search for and exploit the specific blind spots or weaknesses introduced into AI models by this synthetic data, turning the information ecosystem itself into an instrument for mass manipulation.
Europe’s current regulatory framework, anchored by the Digital Services Act (DSA) and the AI Act, is proving insufficient to counter this development. These laws were designed for an earlier generation of technology, focusing primarily on platform compliance, transparency, and the mitigation of risks associated with static content. They currently lack provisions to address coordinated activity by autonomous agents that plan and adapt independently. Even institutional bodies, such as the European Centre for Democratic Resilience, lack the technical mandates and the offensive “red-teaming” capacity necessary to identify and preempt threats generated by these sophisticated, goal-driven systems.
The implications of this defensive gap are severe, extending well beyond minor political annoyance. Autonomous AI-driven attacks are poised to target the EU’s geopolitical cohesion, particularly regarding support for Ukraine, ongoing debates about EU enlargement, and the integrity of future European elections. Moscow is strategically motivated to exploit these divisions, potentially eroding public trust in democratic institutions and the legitimacy of AI itself as citizens realize, or even suspect, that the tools they rely on for information are being subtly weaponized against them.
To survive this shift, the European Union must transition from high-level legislative commitments to concrete operational capabilities. This requires formalizing the role of the Centre for Democratic Resilience with the resources and technical mandate to conduct sustained red-teaming of FIMI (Foreign Information Manipulation and Interference) networks. Furthermore, the implementation of the DSA and AI Act must evolve to target the coordinated behavior of agent networks rather than just the content they produce. Establishing a dedicated funding line for “public-interest AI” tools—focused on detecting autonomous behavior and generating rapid, accurate counter-narratives—will be essential to reducing reliance on external, private-sector providers.
Ultimately, the asymmetry between authoritarian states and open societies leaves Europe vulnerable if it remains tethered to outdated regulatory logic. While existing legal frameworks provide a vital baseline, they are ill-equipped to audit systems capable of planning and modifying their strategies in real time. Unless Brussels shifts its democracy protection instruments, AI regulation, and security policy to account for the realities of agentic operations, it will remain trapped in a perpetual cycle of reacting to campaigns after the damage is already done, rather than actively shaping the security of its information environment.
