Abstract

I argue that the United Kingdom and the European Union are losing an influence contest in which artificial intelligence is used to shape what publics believe and how they then act. Russia and China are using AI for disinformation and propaganda that reduces trust in democratic institutions and widens division. The British and European answer has concentrated on regulation, ethical limits and defence. It has been slow, and it is split across institutions. Without a strategy that uses AI in information operations, and that hardens society against this form of pressure, the disadvantage grows.

Introduction

The contest is already under way, and the published assessment is blunt. Analysts describe Western democracies as “losing the battle of the narrative” on AI (sciencebusiness.net). Britain and the EU member states face AI-supported disinformation while law, ethics and institutional division constrain the reply. If they cannot hold this ground, the expected results are lower public trust, deeper social division, and less influence on decisions taken elsewhere.

The Cognitive Battlespace in the Age of AI

I use “cognitive domain” for the contest over belief and behaviour. In that domain the object of the operation is how people judge and how they act. One formulation of the aim is to “change not only what people think, but how they think and act” (esdc.europa.eu). Sustained, the pressure can fragment a society and reduce its willingness to resist an adversary.

AI raises the scale and the specificity of the effort. Models can tailor a propaganda feed to a person, amplify a false story through a bot network, and produce images and video that are hard to tell from a record of an event. The work can be done at a distance, and it can be aimed at a whole society. NATO’s Allied Command Transformation describes cognitive warfare as activity that affects attitudes and behaviour by influencing cognition, “modifying perceptions of reality” as a standing form of conflict (act.nato.int). Social media platforms and information systems are the means.

In May 2023 an AI-generated image of an explosion at the Pentagon circulated, pushed by bots. Enough people believed it, briefly, that the US stock market dipped. The image was then debunked. The episode is the case a Modern War Institute analysis uses for the “catastrophic potential” of AI-driven disinformation to move events at machine speed (mwi.westpoint.edu). Generative systems can produce convincing false content in volume, and the volume is what makes a public check difficult.

The asymmetry follows from the cost. An attacker can launch a campaign cheaply and at low risk to itself. The defender has to find and answer each new false story. As the volume rises, journalists, fact-checkers and institutions cannot process it at the same rate. A democracy that does not change its information defence loses ground for that reason.

Adversaries’ AI-Driven Influence Operations

Russia and China are the principal state users of AI for influence against European societies. Russia has run information operations against the West for a long time and has added AI to them. Troll farms and bot networks push generated propaganda at the target society. Moscow treats the method as a way to “destroy [the West] from within” without a shot. Expecting to lose a direct war with NATO, the Kremlin funds what the same account calls “cheap and highly effective” cognitive warfare against democratic societies from the inside (kew.org.pl).

Russian military writing calls the technique reflexive control. The operator works on the adversary’s perception until the adversary produces the result the operator wanted. The practice is to fill the information space with tailored disinformation, conspiracy and emotionally charged material, in order to reduce the will to resist. The object is the citizen’s judgement. Emotional reaction, confirmation bias and confusion are the openings. The stated effect is to leave people “unable to properly assess reality”, and therefore less able to decide, including at an election. A democracy divided against itself, and short of trust in its own institutions, is less able to hold a common line against Russian pressure.

The methods are in use. Beyond the false Pentagon image, Russian operators have used generative AI for deepfake video of European politicians, for fabricated news, and for the impersonation of media outlets. Before Western elections, campaigns augmented with AI have aimed to polarise opinion. UK intelligence reporting and independent researchers have found Russian bot accounts using language models to produce messages that are more “human-like and persuasive”. A changed election result is not required for the Kremlin’s gain. Eroded trust in information, and in institutions, is the longer effect.

China is a more recent operator in European information space, and it is building AI into influence work. Chinese military strategy includes “cognitive domain operations”, which combine AI with psychological and cyber warfare. The aim is to shape narratives so that opposition to Chinese interests costs more. China has used bot networks to spread disinformation on the origins of COVID-19 and on the status of Taiwan, and it has generated deepfake news anchors and social-media personas that carry pro-China narratives abroad. NATO analysts report that China describes this warfare as the use of public opinion and psychological manipulation to reach a political result, and that Beijing invests in related technology, including systems that monitor soldiers’ emotions. The European campaign is less open than Russia’s. Its observable purposes are to reduce criticism of Beijing and to divide Western positions.

Authoritarian operators are not bound by the legal and ethical limits democracies accept. They use speed, volume and ambiguity to move before a defence has formed, and they line information operations up with cyber attacks, diplomatic pressure and economic coercion. On the record above, the initiative sits with Russia and China. The UK and the EU are answering a campaign that has already started.

UK and EU Responses: Strategies and Shortcomings

The United Kingdom and the European Union have recognised the problem and have started to build defences against disinformation. The effort so far answers campaigns after they circulate. It is cautious, it is divided across institutions, and it lags the rate at which the campaigns change.

United Kingdom: The Ministry of Defence’s Defence Artificial Intelligence Strategy (2022) warns that “AI could also be used to intensify information operations, disinformation campaigns and fake news,” including by deepfakes and false social-media accounts. The Integrated Operating Concept (2020) treats information operations as part of the answer to false narratives (gov.uk). Britain has stood up units for strategic communications and cyber influence, and it works with NATO on information security.

The Alan Turing Institute’s centre CETaS has published analysis of AI-enabled influence operations in the 2024 UK and European elections. The work identifies deepfake propaganda and AI-generated political smears as emerging threats. It also finds that AI’s effect on those elections was limited, and it flags AI-driven incitement and voter confusion as the serious near-term problems (cetas.turing.ac.uk). Traditional disinformation still does most of the work. The reason to change policy is the rate at which generative tools can scale a campaign that today is still mostly handmade. British discussion has turned to transparency rules for political advertising that uses AI, and to education that helps the public recognise fabricated content.

Critics argue that the British response is split and short of resources. There is no public doctrine for cognitive warfare comparable to the adversary doctrines above. The work is divided among GCHQ, the Army’s 77th Brigade and the Foreign Office, and the parts do not add up to one campaign. Fact-checking and the removal of false accounts have improved. Britain has been unwilling to run information operations that would get ahead of an adversary narrative. Law, ethical limits and the risk of escalation constrain those operations. The result is that a response often starts after the campaign has already reached part of the public.

European Union: The EU has treated disinformation mainly as a problem of election security and of platform duty. It has a Code of Practice on Disinformation, a voluntary arrangement with technology companies, and the East StratCom Task Force, whose EUvsDisinfo project debunks pro-Kremlin claims. After interference in elections and referendums, EU institutions have named Russia as the principal source of disinformation aimed at Europe. The Digital Services Act requires large platforms to assess and reduce risks from false content.

The AI Act reaches manipulation by requiring transparency for deepfakes. The wider EU effort remains organised around defence and norms: regulate the platforms, inform the public, and do not run influence operations. Leaders refuse to blur a reply to propaganda with propaganda of their own, because of the commitment to free expression. An open society that copied authoritarian methods would damage the arrangement it was trying to defend. That limit is real, and it leaves a gap the adversary is organised to use.

A course at the European Security and Defence College has called cognitive warfare an “emerging challenge” for the Union (esdc.europa.eu). Participants discussed tools to detect, deter and mitigate the threat. Outside that specialist setting there is no EU military command for cognitive warfare of the kind that exists for land, sea or cyber. NATO, which includes most EU members, has done more of the conceptual work. NATO’s mandate limits what it will do on the offensive side of information operations.

A RUSI commentary makes the gap specific. NATO’s 2024 strategy update used strong language on the urgency of AI-enabled disinformation, and the same strategy “makes no reference to how AI could be used” for strategic communications or for countering disinformation (rusi.org). The documents emphasise governance standards, public resilience and checking what is true. They do not set out an offensive use of AI. Adversaries who have rejected that restraint keep the initiative.

Why the West Is Losing the AI Influence War

Five features of British and European policy explain the loss of ground.

  • The reply comes after the campaign. Democracies answer disinformation once it has circulated, with fact-checks and diplomatic condemnation. They have not adopted a strategy for establishing the account first. Adversaries publish the operation first.
  • Law and ethics rule out a matching method. A duty to truthfulness, to law and to civil liberty rules out a government deception campaign that would be illegal or politically unacceptable at home. Matching Russian or Chinese scale by the same means would damage the standard being defended. Authoritarian governments do not carry that constraint.
  • Authority is split. The threat crosses domestic agencies, the EU and NATO, and no single command directs the information reply. Adversary messaging is more often run from one set of instructions, which makes it faster and more consistent.
  • Regulation has outrun capability. EU rules on AI, social media and data build constraints slowly. Rules can reduce some harmful content. They do not stop a determined state, and they do not create British or European tools for strategic communication or for a fast counter-message. The investment gap leaves the technical initiative with the opponent.
  • Recognition came late. Western doctrine long ranked physical and cyber threats above information warfare. A campaign against judgement, conducted in peacetime, was not a standard planning case. Before 2016, few plans assumed that online influence could bear on a major vote. The lag gave adversaries time to build the practice.

Awareness has risen, and the countermeasures have not caught the campaigns. Conspiracy content that goes unanswered, division driven through disinformation, and doubt cast on institutions each reduce the capacity for a common response. NATO officials warn that information-warfare threats “must neither be overlooked nor underestimated” as AI spreads. The present effort still begins after the false content is circulating.

Conclusion and Implications

If the UK, the EU and democracies in a similar position do not adapt, the loss will not arrive as a surrender. It will arrive as weaker resilience: deeper division, less trust in media and in government, and adversary narratives that settle in. That erosion reduces the political will required for a response to more conventional threats. One analysis states the cost of inaction as a “strategic imbalance favouring adversaries”, in which false narratives reduce public trust and, left unanswered, damage democratic institutions.

Reversing the imbalance means treating the cognitive domain as a problem of national security. Four requirements follow.

  • Tools for defence and for a reply. Invest in detection and in correction, and in the distribution of true counter-narratives against authoritarian propaganda. The ethical rules for those operations have to be written down. Refusing to use AI in the information contest leaves the field to the adversary.
  • A public that can recognise the method. Education in how media works, a public account of the threat, and independent journalism reduce the effect of a manipulated feed. Recognition is the defence that scales past any single fact-check. The aim is that citizens treat digital information with enough scepticism to limit the damage from fabricated content.
  • One front rather than several. NATO’s work on cognitive warfare needs matching UK and EU arrangements for sharing intelligence on campaigns and for aligning replies. A common front denies the adversary the chance to play one democracy against another.
  • Rules that can be changed at the speed of the tool. Faster duties on platforms, election law that covers content generated by AI, and narrow offences for the most dangerous deceptive media, including media that incites violence. Free speech is the constraint. The balance has to be reset in law as the technology changes, not left as an unstated preference.

The cognitive domain is now a permanent part of international security. An open society can defend it. The defence is the use of AI for detection and for a truthful counter-narrative, inside rules a democracy can publish. The alternative is to leave the initiative with those who, in the phrase this literature uses, “attack the democratic pillars of the West” through the manipulation of information. The result follows from whether the state can act before the campaign has done its work.

Bibliography

1. NATO Allied Command Transformation. “Cognitive Warfare.” NATO ACT, Norfolk VA.

2. Bryc, Agnieszka. “Destroy from within: Russia’s cognitive warfare on EU democracy.” Kolegium Europy Wschodniej, 27 Nov 2024.

3. European Security & Defence College (ESDC). “Cognitive warfare in the new international competition: an emerging challenge for the EU,” 28 May 2024.

4. Williams, Cameron (Modern War Institute). “Persuade, Change, and Influence with AI: Leveraging Artificial Intelligence in the Information Environment.” Modern War Institute at West Point, 14 Nov 2023.

5. UK Ministry of Defence. Defence Artificial Intelligence Strategy, June 2022. UK

6. Fitz-Gerald, Ann M., and Halyna Padalko (RUSI). “The Need for a Strategic Approach to Disinformation and AI-Driven Threats.” RUSI Commentary, 25 July 2024.

7. Science Business News. “EU is ‘losing the narrative battle’ over AI Act, says UN adviser,” 05 Dec 2024.