The landscape of digital deception is undergoing a seismic shift that renders our current defenses almost obsolete. While we have spent the last few years playing a high-stakes game of “whack-a-mole” against basic artificial intelligence—the kind that spits out simple, albeit prolific, falsehoods—a much more dangerous evolution is underway. We are moving from the era of generative AI, which follows human prompts, into the age of “agentic” systems. These new tools are goal-oriented; you can give them a target objective, and they will independently navigate the web, test tactics, monitor their own progress, and pivot strategies without needing a human hand at the wheel. For adversaries like Russia, this represents a quantum leap in influence operations, turning decentralized, relentless digital campaigns into a persistent, autonomous force rather than just a collection of spam.
This transition poses a unique threat because it attacks the very “supply chain” of our information ecosystem. We have already seen Russian-backed actors flood the internet with millions of low-quality, biased articles specifically designed to be scraped by mainstream AI models. The result is that even the most reputable Western chatbots have been caught repeating Kremlin-aligned narratives because they were trained on a poisoned web. By targeting the informational base that AI platforms use to learn, bad actors are effectively turning our own technological advancements into instruments of mass manipulation. It is no longer just about a fake post on social media; it is about polluting the underlying intelligence of the digital tools the public uses every day.
The current European regulatory framework, including the Digital Services Act and the AI Act, was built for an earlier generation of technology. These laws are heavily focused on transparency, content labeling, and platform accountability—all noble goals, yet they fail to address the nuance of coordinated, autonomous agent networks. Current policy treats AI as a static interface, missing the fact that agentic systems are active, planning entities capable of exploiting systemic blind spots in real time. Because Brussels lacks the technical mandate or the specialized “red-teaming” capacity to stress-test these autonomous threats, lawmakers are essentially trying to regulate the behavior of a fast-moving, adaptive entity using a rulebook written for static software.
This defensive gap carries heavy political consequences that extend far beyond mere annoyance. By deploying relentless waves of autonomous operations, adversaries can systematically fracture support for essential geopolitical goals, such as aid for Ukraine or EU enlargement, precisely at the moments when European cohesion is most fragile. The danger is not just that disinformation succeeds; it is that the public may lose faith in the integrity of AI itself. As people become aware—or even just suspicious—that their chatbots are being subtly guided by foreign handlers, the baseline of societal trust erodes. In this asymmetric conflict, authoritarian actors operate with few internal constraints, while democratic societies must find a way to defend their discourse without sacrificing the open platforms that define them.
To bridge this divide, the European Union must transition from high-level, abstract commitments to granular, operational reality. This means evolving the European Centre for Democratic Resilience into an active, technical powerhouse capable of sustained, real-time testing of agentic threats. We need a permanent, dedicated “red-teaming” operation that can detect coordinated agent behavior before it hits the mainstream. Simply waiting for a platform to remove a piece of content is no longer a viable security strategy. Regulation must shift its focus from the output—the synthetic text or deepfake—to the infrastructure: the networks of agents that coordinate these campaigns to manipulate human perception and exploit systemic bias.
Ultimately, the goal must be building sovereign resilience. We cannot rely solely on the slow movement of traditional research and development to counter a threat that evolves in seconds. Europe needs dedicated, agile funding for public-interest AI tools designed to detect synthetic interactions and neutralize orchestrated manipulation. We must move beyond basic media literacy toward a deeper public understanding of how these autonomous agents influence our digital interactions. If Brussels does not rapidly align its security policy, legislative frameworks, and technological investments to match the reality of agentic operations, it will remain trapped in a cycle of permanent reaction, always one step behind an adversary that is no longer just writing propaganda, but automating the erosion of our democracy.