Home Economy THE SYNTHETIC FRONTIER: ARTIFICIAL INTELLIGENCE AS A SYSTEMIC DRIVER OF GLOBAL DISINFORMATION...

THE SYNTHETIC FRONTIER: ARTIFICIAL INTELLIGENCE AS A SYSTEMIC DRIVER OF GLOBAL DISINFORMATION AND GEOPOLITICAL INSTABILITY (2025–2030)

Dicembre 1, 2025

242

EXECUTIVE NARRATIVE ABSTRACT: THE ERA OF EPISTEMIC INSECURITY

The global information environment has entered a period of critical instability, defined not merely by the presence of falsehoods but by the industrial-scale automation of reality distortion. The convergence of Generative Artificial Intelligence (GenAI), autonomous distribution networks, and state-sponsored hybrid warfare doctrines has elevated information integrity from a peripheral communications issue to a Tier-1 global security threat. This report, synthesizing verified data from over 140 government, intergovernmental, and academic sources—including the United Nations, NATO Strategic Communications Centre of Excellence (StratCom COE), the European External Action Service (EEAS), and the World Economic Forum (WEF)—establishes that the 2024–2025 period marks the definitive “phase transition” where synthetic media became a primary vector for geopolitical and economic destabilization.

The statistical evidence for this shift is absolute. The World Economic Forum’s Global Risks Report 2025 explicitly identifies misinformation and disinformation as the single most severe global risk over a two-year horizon, surpassing traditionally dominant threats such as extreme weather events, economic downturns, and interstate armed conflict.¹ This ranking is not a speculative forecast but a reflection of current operational realities, derived from the Global Risks Perception Survey (GRPS) which aggregates insights from over 1,100 experts across academia, business, and government.³ The consensus is stark: 80% of experts indicate that disinformation is already occurring at a crisis level, with nations reporting they feel “least prepared” to address this specific threat compared to any other global risk.³

This escalation is driven by a fundamental change in the economics of influence. The barrier to entry for high-sophistication disinformation has collapsed. What previously required a well-funded intelligence agency and a dedicated “troll farm” of human operators can now be executed by Agentic AI systems capable of autonomous narrative generation, content production, and cross-platform dissemination. The United Nations Development Coordination Office characterizes this not as a technical challenge but as a “crisis of trust” capable of unraveling the social fabric itself.³ The UN Global Principles for Information Integrity, launched in June 2024, explicitly link the proliferation of AI-generated content (AIGC) to the erosion of human rights, warning that in settings of instability, these tools can “tip societies into violence”.³

The geopolitical landscape has fractured under this pressure. While the European Union has fortified its defenses through the Digital Services Act (DSA) and the Foreign Information Manipulation and Interference (FIMI) framework—mapping over 505 distinct incidents and 38,000 disinformation channels in a single year—the United States has dismantled key counter-disinformation infrastructure.⁶ The closure of the U.S. State Department’s Global Engagement Center (GEC) in April 2025 signals a profound transatlantic divergence, leaving a vacuum in the global monitoring architecture that hostile actors are rapidly filling.⁸

Simultaneously, the weaponization of identity has breached the corporate and financial sectors with devastating efficacy. The verified “Arup Incident” of 2024, where a multinational firm lost $25.5 million to a deepfake video conference scam, demonstrated that “live” interaction is no longer a guarantor of human identity.¹⁰ With banking fraud losses projected to reach $40 billion by 2027 due to GenAI-enabled attacks, the global economy faces a crisis of verification that threatens the foundations of digital commerce.¹¹

This report provides an exhaustive, empirically grounded analysis of these phenomena. Rejecting speculative futurism, it dissects confirmed campaigns in Taiwan, Ukraine, and the Euro-Atlantic sphere to model the trajectory of AI-driven hybrid warfare through 2030.

The Cognitive Limits of Warfare: U.S. Army Professional Military Education, Artificial Intelligence Integration and Human Adaptation in the Accelerating Speed of Conflict (2025)

1. EXECUTIVE NARRATIVE ABSTRACT: THE ERA OF EPISTEMIC INSECURITY
2. CHAPTER 1: THE NEW GLOBAL RISK CALCULUS AND THE FAILURE OF CONTAINMENT
2.1 The Verified Escalation of Threat Levels (WEF & UN Risk Analysis)
2.2 The “Confusion” Doctrine: A Strategic Shift (NATO StratCom Analysis)
2.3 The Generational and Societal Divide
3. CHAPTER 2: THE GEOPOLITICS OF SYNTHETIC REALITY AND STATE ACTOR CONVERGENCE
3.1 The Russian Federation: The “Firehose” and Deepfake Diplomacy (Case Study: Olena Zelenska)
3.2 The People’s Republic of China: “Spamouflage” and Narrative Engineering
3.3 The Sino-Russian Convergence: Operational Synchronization
4. CHAPTER 3: ELECTORAL BATTLEGROUNDS AND THE TRANSATLANTIC POLICY SPLIT
4.1 The Taiwan 2024 Case Study: The AI Crucible
4.2 The United States: The Dismantling of the GEC and “Free Speech” Frameworks
4.3 The European Union: The Fortress Strategy (DSA & FIMI Matrix)
5. CHAPTER 4: THE ECONOMICS OF DECEPTION AND THE $40 BILLION FRAUD HORIZON
5.1 The Arup Incident: Anatomy of a $25 Million Deepfake
5.2 The Quantitative Scale of the Crisis (Global Fraud Rates)
5.3 The Collapse of Identity Verification (KYC) and “Liveness” Detection
6. CHAPTER 5: COGNITIVE WARFARE AND THE PSYCHOLOGY OF SYNTHETIC REALITY
6.1 The Illusion of Detection: Empirical Failure (Science & PNAS Studies)
6.2 The “Liar’s Dividend” and the Erosion of Trust
6.3 AI Persuasion Capabilities vs. Human Influence
6.4 Psychological Violence and Gendered Impacts (NCII)
7. CHAPTER 6: FUTURE HORIZONS (2026–2030) AND THE RISE OF AGENTIC AI
7.1 Agentic AI and Automated Influence Operations
7.2 The “Splinternet” of Trust: Fragmented Regulatory Zones
7.3 Conclusion: Structuring Resilience in the Synthetic Age
8. REFERENCES

THE NEW GLOBAL RISK CALCULUS AND THE FAILURE OF CONTAINMENT

The assertion that disinformation is a manageable externality of the digital age has been definitively dismantled by the empirical realities of 2024 and 2025. The global strategic community has realigned its threat perception matrices, acknowledging that the integrity of information is now a foundational requirement for economic stability, national security, and democratic continuity.

The Verified Escalation of Threat Levels

The trajectory of risk perception regarding AI and disinformation has been vertical. The World Economic Forum (WEF) Global Risks Report 2025 presents a watershed moment in risk analysis. For the first time, “misinformation and disinformation” have ascended to the number one rank among global risks for the short-term (two-year) outlook.¹ This displaces “cost-of-living crisis,” “cyber insecurity,” and “armed conflict” from the top spot. The report emphasizes that this risk is “interconnected” and “poly-crises” inducing; it does not exist in a vacuum but acts as a catalyst that accelerates societal polarization and weakens governance structures, making all other risks harder to manage.²

The United Nations echoes this urgency. The UN Global Principles for Information Integrity, launched in June 2024, were a direct response to the “harm caused by the spread of online hate and lies”.¹³ Secretary-General António Guterres highlighted that advances in AI have created a vector for these harms that is unprecedented in speed and scale. The UN Development Coordination Office notes that in a world shaken by climate shocks and inequality, disinformation is the “least visible” but perhaps most dangerous threat because it “corrodes the norms of debate and science-backed evidence” essential for solving those very problems.³

The data supports the UN’s grim assessment. A survey of over 1,100 experts from 136 countries found that 80% feel disinformation is already happening at a dangerous scale, and it is the risk for which they feel their national institutions are least prepared.³ This “preparedness gap” is the critical vulnerability that GenAI exploits.

The “Confusion” Doctrine: A Strategic Shift

Why has containment failed? The NATO Strategic Communications Centre of Excellence (StratCom COE) provides a critical military-strategic perspective. The Centre’s analysis for 2025 indicates a shift in adversarial strategy from “persuasion” to “confusion.” In the past, propaganda aimed to convince a target audience of a specific alternative truth (e.g., “Our system is better than yours”). The new doctrine, empowered by AI, aims to “muddy the waters” with so much contradictory information that the target audience becomes incapable of distinguishing fact from fiction.¹⁴

This “confusion” strategy is computationally efficient. It is far easier to generate 10,000 conflicting AI-generated narratives than to construct one coherent, persuasive argument. The goal is to induce “decision paralysis” and “cognitive fatigue” in the adversary’s population. NATO StratCom COE identifies this as a core component of hostile information operations, where the sheer volume of “noise” prevents democratic consensus formation.¹⁴

The volume of these operations is quantifiable. The European External Action Service (EEAS) detected and analyzed 505 FIMI (Foreign Information Manipulation and Interference) incidents between November 2023 and November 2024.⁷ These were not random acts of trolling but coordinated campaigns involving over 38,000 unique channels across 25 different platforms.⁷ The EEAS report recorded over 68,000 observables (pieces of content), confirming that Europe is facing a sustained, high-intensity bombardment of information warfare.⁷

The Generational and Societal Divide

The impact of this technological shift is unevenly distributed, creating new forms of social stratification. The United Nations highlights a growing “digital literacy gap” that is profoundly generational. A poignant example cited involves an elderly individual unable to distinguish an AI-generated video of an elephant from reality. While a “fake elephant” may seem harmless, the psychological impact of being “tricked” and the realization that “nothing is real” erodes the epistemic confidence of older demographics.³

However, the WEF warns that this is not merely a problem of the elderly. Younger generations, while digitally native, are being targeted by “micro-targeted” AI campaigns that exploit their specific grievances and cultural touchpoints. The WEF Global Risks Report notes that disinformation is a leading mechanism for foreign entities to “affect voter intentions” and “sow doubt among the general public worldwide about what is happening in conflict zones”.¹²

The societal implications are profound. The UN warns that in settings with increased instability, AI disinformation can “tip societies into violence”.³ This kinetic effect—where digital lies lead to physical harm—is the ultimate manifestation of the threat. The UNESCO recommendations and the UN Global Principles emphasize that governments must now provide resources for “media and digital literacy campaigns” that are as robust as their public health or defense initiatives.¹⁶

EXCLUSIVE REPORT – Impact of Military Artificial Intelligence on Nuclear Escalation Risk: Geopolitical, Technological and Strategic Dimensions in 2025

THE GEOPOLITICS OF SYNTHETIC REALITY AND STATE ACTOR CONVERGENCE

The era of the “lone hacker” is effectively over. The 2024–2025 period is defined by the industrialization of influence operations by major state powers, specifically the Russian Federation and the People’s Republic of China (PRC). These actors have integrated AI into their strategic doctrines, creating a convergence of tactics that threatens the Euro-Atlantic and Indo-Pacific security architectures.

The Russian Federation: The “Firehose” and Deepfake Diplomacy

Russia remains the primary architect of aggressive, chaotic disinformation. The NATO StratCom COE analysis reveals that Kremlin-aligned messaging bursts are roughly twice as frequent as their pro-Western counterparts and exhibit tighter synchronization across platforms.¹⁵ The Russian strategy leverages AI to overwhelm the information space with “white noise” and manufactured scandals.

A definitive case study of this capability is the Olena Zelenska “Bugatti” Deepfake. In 2024, a deepfake video circulated widely on social media, falsely accusing the First Lady of Ukraine, Olena Zelenska, of purchasing a Bugatti Tourbillon sports car worth $4.8 million using U.S. aid money.¹⁸ The operation was sophisticated:

The Asset: An AI-generated deepfake video of a supposed “whistleblower“—a car dealership employee—detailing the purchase.¹⁹
The Infrastructure: The video was hosted on a fake verification website named “Veritée Caché,” designed to mimic a legitimate French investigative journalism outlet.¹⁹
The Amplification: Russian state-aligned networks and bot farms rapidly amplified the story, targeting U.S. and European taxpayers to undermine support for Ukraine aid packages.²⁰

Although the story was debunked by Full Fact, CBS News, and cybersecurity experts who proved the invoice was forged and the “employee” did not exist, the narrative achieved millions of views before corrections could take hold.¹⁸ This illustrates the asymmetry of AI warfare: the cost of generation is near-zero, while the cost of refutation is high.

Verified reports from the Atlantic Council’s DFRLab and Europol also detail the deployment of deepfake videos depicting Ukrainian President Volodymyr Zelenskyy surrendering.²¹ While early attempts in 2022 were technically crude, the 2024 wave demonstrated a marked increase in audio-visual fidelity, utilizing advanced lip-syncing and voice cloning technologies that are increasingly difficult for automated filters to detect.²²

The People’s Republic of China: “Spamouflage” and Narrative Engineering

In contrast to Russia’s “firehose of falsehood,” the PRC employs a strategy of “narrative engineering” focused on reshaping global perceptions of China and discrediting Western democracy. Graphika, a leading intelligence firm, has tracked the evolution of the Spamouflage network—a state-linked influence operation. By late 2024 and early 2025, Spamouflage had evolved from “spammy” posts to highly targeted, persona-based operations.²³

Intelligence reports confirm that Chinese actors utilized AI to impersonate U.S. voters on platforms like X (formerly Twitter) and TikTok. These AI-generated personas did not just spread pro-China slogans; they simulated “frustration” with American politics, adopting the vernacular and grievances of specific U.S. demographics to depress voter turnout and exacerbate domestic divisiveness.²⁴

A disturbing escalation occurred in Europe, where Spamouflage assets posed as the Spain-based non-profit Safeguard Defenders. The network spread online calls for the overthrow of the Spanish government following deadly floods in Valencia—a “false flag” operation designed to discredit the actual Safeguard Defenders organization, which is a vocal critic of Beijing’s human rights record.²³ This represents a new frontier: the use of AI to frame human rights groups for inciting insurrection.

Furthermore, the PRC utilizes AI to deflect attribution. Following accusations of Chinese hacking by Japan in early 2025, Chinese state media and AI-amplified networks disseminated AI-generated cartoons and narratives casting Tokyo as an agent of U.S. disinformation. This “reverse-accusation” strategy muddies the attribution waters, making it difficult for the public to discern the aggressor.²³

The Sino-Russian Convergence

Perhaps the most alarming trend identified by the Center for European Policy Analysis (CEPA) and the EEAS is the operational convergence of Russian and Chinese tactics. While their long-term strategic goals may differ, their methods in the information domain are aligning.

Both powers are utilizing Generative AI to produce content that reinforces anti-Western narratives, particularly regarding the war in Ukraine and NATO’s global role. The EEAS 3rd Report on FIMI Threats highlights a “functional convergence” where Russian and Chinese assets amplify each other’s content during major geopolitical events.²⁵ For example, during the 1,000-day mark of the invasion of Ukraine, both ecosystems synchronized narratives blaming NATO for “escalating” the conflict.²⁵ This creates a “hybrid storm,” where the combined weight of two state-backed AI ecosystems can dominate global algorithmic trends, drowning out democratic messaging.²⁶

Navigating the Integration of Artificial Intelligence in Nuclear Enterprises: Balancing Efficiency, Stability and Human Oversight

ELECTORAL BATTLEGROUNDS AND THE TRANSATLANTIC POLICY SPLIT

The years 2024 and 2025 were characterized by a historic “super-cycle” of elections, serving as a global crucible for AI-driven interference. The responses of democratic nations to this threat have diverged sharply, creating a fragmented global defense posture.

The Taiwan 2024 Case Study: The AI Crucible

The January 2024 presidential election in Taiwan serves as the premier case study for modern AI-driven disinformation. The island democracy faced an unprecedented barrage of cognitive warfare from the PRC, utilizing what researchers at Dartmouth and the Thomson Foundation described as a sophisticated, multimodal campaign.²⁷

A key vector was the use of deepfake audio and video. In the critical weeks leading up to the election, AI-generated content circulated depicting the Democratic Progressive Party (DPP) candidate, Lai Ching-te, discussing “illegal” activities. One specific deepfake featured a synthetic voice of Lai referring to himself as “immoral Lai” and claiming his party was fraught with scandals.²⁹ The Taiwan Investigation Bureau had to publicly intervene, identifying the audio recordings as deepfakes, marking one of the first times a government agency formally attributed a deepfake attack during an active election cycle.²⁸

Crucially, the disinformation did not always attack candidates directly. A study published in the Harvard Kennedy School Misinformation Review found that a significant portion of the narratives targeted US-Taiwan relations, attempting to sow “skepticism” about American reliability.²⁷ This “US Skepticism” narrative was amplified by AI-generated avatars on platforms like Line, PTT, and Facebook.²⁷ Despite this, Taiwan’s resilience was notable. The deployment of indigenous AI tools like the Trustworthy AI Dialogue Engine (TAIDE) helped counter the narrative flood, and the election of Lai Ching-te was viewed as a victory for democratic resilience.²⁸

The United States: The Dismantling of the GEC

In a move that stunned allies, the United States radically altered its counter-disinformation posture in 2025. Following a change in administration, Secretary of State Marco Rubio directed the closure of the Global Engagement Center (GEC) in April 2025.⁸ The GEC had been the primary U.S. interagency hub for coordinating efforts to expose and counter foreign state propaganda.³²

Official statements cited that the framework had “devolved into tools for political censorship” and directed the cessation of all associated instruments.⁸ This was aligned with an executive order on “Restoring Freedom of Speech”.⁸ While framed as a civil liberties victory domestically, the closure left a significant vacuum in the global intelligence architecture. NATO allies expressed concern that the “intelligence gap” left by the GEC’s dissolution would hinder collective detection capabilities.³³ The U.S. effectively retreated from the field as a centralized coordinator of global “pre-bunking” operations, shifting the burden of monitoring Russian and Chinese influence to European agencies and the private sector.³³

The European Union: The Fortress Strategy

While the U.S. retreated, the European Union fortified its position. The European Commission and EEAS aggressively implemented the Digital Services Act (DSA), which mandates that large platforms mitigate systemic risks, including disinformation.³⁴ The EEAS 3rd Report on FIMI Threats established a “FIMI Exposure Matrix,” a standardized framework for mapping the supply chain of disinformation.⁶

The EU’s approach is highly institutionalized. The EUvsDisinfo project continues to catalog thousands of pro-Kremlin cases.³⁶ In 2025, the EU focused on mapping the “digital infrastructure” of foreign actors—identifying the specific servers, bot networks, and “state-aligned” proxies used to disseminate FIMI.⁶ This divergence creates a geopolitical vulnerability: the transatlantic alliance now operates with decoupled strategies—Europe relying on state regulation and centralized monitoring, and the U.S. relying on a decentralized “marketplace of ideas.”

The Integration of Artificial Intelligence in the Sukhoi Su-57: Geopolitical, Economic and Technological Implications for Global Aerospace Defense

THE ECONOMICS OF DECEPTION AND THE $40 BILLION FRAUD HORIZON

The weaponization of AI has migrated from the political sphere to the financial sector, evolving from a reputational threat into a direct instrument of massive capital theft. The 2024–2025 period witnessed the collapse of traditional identity verification under the weight of deepfake technology.

The Arup Incident: Anatomy of a $25 Million Deepfake

The most significant verified financial loss to date occurred in early 2024, involving the multinational engineering firm Arup. An employee at the firm’s Hong Kong office was targeted by a highly sophisticated deepfake operation.

The Sequence of the Attack:

Initial Contact: The employee received a message purportedly from the company’s UK-based Chief Financial Officer (CFO) regarding a confidential transaction.¹⁰
The “Liveness” Trap: When the employee expressed skepticism, the fraudsters invited them to a video conference call.
The Synthetic Meeting: On the call, the employee saw the CFO and several other familiar colleagues. Crucially, every participant on the call except the victim was a deepfake—AI-generated avatars indistinguishable from the real individuals in real-time.¹⁰
The Execution: Convinced by the visual and auditory evidence, the employee authorized 15 transfers totaling $25.5 million to fraudulent accounts.¹⁰

This incident, confirmed by Sumsub, Deloitte, and major financial news outlets, represents a watershed moment: “The Arup Incident.” It demonstrated that the “uncanny valley” has been crossed and that live video interaction is no longer a guarantor of identity.³⁸

The Quantitative Scale of the Crisis

The banking and finance sectors are facing a tsunami of AI fraud. Deloitte’s Center for Financial Services projects that Generative AI could enable fraud losses to reach $40 billion in the United States alone by 2027, growing at a compound annual rate of 32% from $12.3 billion in 2023.¹¹

Key indicators of this surge include:

Deepfake Surge: In North America, deepfake fraud incidents surged by 1,740% between 2022 and 2023.¹⁰
Cost Per Incident: In 2024, the average cost of a deepfake-related fraud incident for businesses was approximately $500,000, with large enterprises facing losses up to $680,000 per incident.³⁹
Global Rates: The Sumsub Identity Fraud Report 2025 indicates a global identity fraud rate of 2.2%. While the overall volume of simple attacks fluctuated, the sophistication skyrocketed, with a 180% increase in complex, AI-enabled fraud attempts compared to the previous year.⁴¹

The Collapse of Identity Verification (KYC)

The core infrastructure of modern finance—Know Your Customer (KYC)—is under siege. Criminals are using deepfakes to bypass biometric verification systems. Entrust reports a 244% surge in digital document forgeries and notes that deepfake attacks now occur every five minutes globally.⁴²

The democratization of these tools is absolute. For as little as $20, bad actors can purchase software on the dark web to create convincing fake identities.¹¹ This renders traditional static identity checks obsolete, forcing institutions to adopt “liveness” detection (e.g., asking a user to blink or turn their head). However, AI agents are now capable of passing these liveness checks in real-time, creating an escalating “arms race” between AI generation and AI detection.⁴³ The IMF has noted that cyber-related losses in the financial sector have quadrupled since 2017, largely driven by these new technologies.⁴⁴

COGNITIVE WARFARE AND THE PSYCHOLOGY OF SYNTHETIC REALITY

The effectiveness of AI-driven disinformation lies in its ability to exploit specific vulnerabilities in human cognition. Recent academic research challenges the comforting notion that humans can reliably detect machine-generated falsehoods.

The Illusion of Detection: Empirical Failure

Empirical studies published in Science and PNAS (Proceedings of the National Academy of Sciences) reveal a disturbing reality: humans are poor detectors of deepfakes, and confidence is not correlated with accuracy.

A study by Nightingale et al. (2023) in the Journal of Cybersecurity found that human accuracy in detecting deepfake faces is unreliable, and specialized training often fails to significantly improve detection rates.⁴⁵ Furthermore, a study in PNAS demonstrated that ordinary humans perform only in the range of leading machine learning models when trying to flag misinformation.⁴⁷ The study found that for every additional 10 seconds a participant spent analyzing a video, their accuracy actually decreased, suggesting that over-analysis leads to confusion rather than clarity.⁴⁷

The “Liar’s Dividend”

The existence of deepfakes creates a secondary effect known as the “Liar’s Dividend.” The mere awareness that deepfakes exist causes the public to doubt authentic information. Research indicates that the presence of deepfakes makes people 37% more likely to distrust legitimate news videos, regardless of the source’s credibility.⁴⁸ This aligns with the NATO StratCom COE‘s observation of “confusion” as a strategic goal: if nothing can be trusted, then truth loses its power to hold power accountable.

AI Persuasion Capabilities

Are AI agents more persuasive than humans? A 2024 PNAS study titled “Testing theories of political persuasion using AI” provides a quantitative answer. The study found that LLM-generated persuasive messages resulted in significant attitude changes across all treatment groups, shifting opinions by an average of 2.5 to 4 percentage points.⁴⁹

Crucially, the study found that the AI was equally persuasive across different strategies (e.g., “generic” vs. “customized” arguments). This implies that the danger of AI is not necessarily that it creates a “super-persuasive” argument, but that it is scalable. An AI agent can engage millions of users simultaneously with competent, effective persuasion—a capability no human propaganda team can match in volume.⁴⁹ However, the study also noted a limitation: while AI could moderate opinions, it did not significantly increase “democratic reciprocity”—the willingness to respect opposing views.⁴⁹

Psychological Violence and Gendered Impacts

The impact of AI disinformation is also deeply personal and gendered. UN Women reports that AI-facilitated violence, such as non-consensual deepfake pornography (NCII), has surged. A global survey found that 85% of women online have witnessed digital violence against others.⁵⁰ This form of “image-based sexual abuse” silences victims, forcing them off digital platforms and causing severe psychological, professional, and financial harm. The speed and anonymity of AI tools allow perpetrators to inflict this damage with impunity, creating a “perfect storm” of gendered disinformation that legal systems are struggling to address.⁵⁰

FUTURE HORIZONS (2026–2030) AND THE RISE OF AGENTIC AI

As we look toward the latter half of the decade, the trajectory of AI disinformation points toward full automation and the emergence of “Agentic” threats that operate independently of human handlers.

Agentic AI and Automated Influence

The next frontier is Agentic AI—autonomous systems that can plan, execute, and adapt influence campaigns without human intervention. The CSET (Center for Security and Emerging Technology) and Google Cloud Cybersecurity Forecast 2026 predict that threat actors will move from using AI as a tool to using it as the “norm“.⁵¹

These agents will be capable of “vibe coding”—creating self-sustaining bot networks that can argue, produce memes, and generate “evidence” in real-time response to breaking news.⁵³ Salesforce executives predict that by 2026, “trusted agentic AI” will be a key differentiator, but the risk of “workslop”—low-quality, hallucinated AI noise—will require businesses to spend hours auditing the very agents meant to save them time.⁵³

The “Splinternet” of Trust

The policy divergence between the US, EU, and China suggests a future where the internet is fragmented into zones of differing “truth.”

The EU Zone: Likely to continue building a “digital fortress” with strict liability for platforms, watermarking requirements, and centralized FIMI monitoring.⁶
The US Zone: Absent a federal agency like the GEC, the US may rely on private sector detection and individual resilience, potentially leaving it more exposed to foreign influence operations but maintaining a more open, less regulated information environment.³³
The Global South: Nations like Brazil, India, and Indonesia will remain primary battlegrounds. Ipsos polling indicates that 77% of citizens in these regions view disinformation as a top threat.⁵⁴ Without the resources of the EU or US, these nations may face the brunt of AI-driven election interference, as seen in the broader Global South trends.⁵⁵

Splinternet is the fragmentation of the internet into separate, siloed networks, often due to government regulations, geopolitical tensions, and differing national interests. This leads to different versions of the internet with distinct rules and limitations, where access to content, services, and information can vary significantly by country. China, Russia, and Iran are often cited as examples of countries creating their own versions of the internet, sometimes referred to as the “Great Firewall” or “Runet”.

Key characteristics and drivers

Siloed cyberspace: The internet is divided into competing, closed cyber-geographies with their own rules and standards.
Geopolitical influences: Authoritarian regimes drive fragmentation by controlling information access, while other countries are motivated by concerns like data privacy, national security, and data sovereignty.
Technological and commercial factors: Export controls on technology (like semiconductors) and bans on foreign hardware also contribute to the splintering effect. The rise of different platforms and incompatible systems can also create barriers.
Varying layers of fragmentation: Splinternet can be seen across different levels of the internet’s infrastructure:
- Physical layer: Different infrastructure, like data centers and cables.
- Logical layer: Varying protocols, languages, and information systems.
- Semantic layer: The applications and content users interact with, which can differ drastically depending on location.
Impact on users: A splinternet can limit access to information and services, as seen in countries with strict internet controls.

Examples of the splinternet in action

China’s “Great Firewall”: Blocks many foreign websites and services like Google, Facebook, and Twitter, with a heavily censored version of the internet available to its citizens.
Russia’s “Runet”: Russia has moved to create its own “sovereign” version of the internet, with government control over its infrastructure.
Iran: The country has an “internet cut off” in certain areas during protests, and users can be imprisoned for online content that doesn’t align with government terms.
Data privacy differences: The same app can have different policies in different regions, like the 2021 WhatsApp privacy policy update which differed for users in Europe versus India due to the GDPR.

Conclusion

The data from 2024 and 2025 is unequivocal: AI has ceased to be a future threat and is now a present operational reality. The cost of inaction is measurable in billions of dollars lost to fraud and the tangible erosion of democratic institutions. As the WEF warns, the risk is not just that we will believe lies, but that the very concept of a shared reality—essential for global cooperation on climate, conflict, and economy—will disintegrate. The response requires not just “fact-checking,” but a fundamental restructuring of digital governance, identifying the “malicious design” of systems, and building societal resilience against the inevitable confusion of the synthetic age.

Table 1: Comparative Disinformation Tactics (2024-2025)

Actor	Primary Objective	Key Tactic	Notable Incident	Source
Russia	Confusion / Chaos	“Firehose of falsehood,” deepfakes	Fake Zelenskyy surrender; Olena Zelenska Bugatti hoax	¹⁴
China (PRC)	Narrative Control	“Spamouflage,” impersonation	Impersonating US voters; Fake Safeguard Defenders campaign	²³
Cyber-Criminals	Financial Theft	Executive Impersonation (CEO Fraud)	Arup Incident ($25.5M loss)	¹⁰
Domestic Actors	Political Polarization	Partisan amplification	Attacks on US-Taiwan relations; US GEC closure debate	⁸

Table 2: Projected Financial Impact of AI Fraud

Metric	Value/Projection	Context	Source
Global Identity Fraud Rate	2.2% (2025)	Down slightly in volume, up massively in sophistication	⁴¹
Deepfake Fraud Growth	1,740% (2022-2023)	North American surge in incidents	³⁸
Banking Fraud Losses	$40 Billion (2027)	Projected US losses due to GenAI fraud	¹¹
Avg. Cost per Incident	~$500,000 (2024)	Cost to businesses per deepfake attack	⁴⁰

References

World Economic Forum. Global Risks Report 2025. Cologny (Switzerland): World Economic Forum; 2025 Jan 15 [Internet]. Available from: https://www.weforum.org/publications/global-risks-report-2025/ Accessed: 2025 Nov 29.
World Economic Forum. Global Risks Report 2025: Digest. Cologny (Switzerland): World Economic Forum; 2025 [Internet]. Available from: https://www.weforum.org/publications/global-risks-report-2025/digest/ Accessed: 2025 Nov 29.
United Nations Development Coordination Office. Disinformation is a global risk. So why are we still treating it like a tech problem?. New York: United Nations; 2024 [Internet]. Available from: https://un-dco.org/stories/disinformation-global-risk-so-why-are-we-still-treating-it-tech-problem Accessed: 2025 Nov 29.
World Economic Forum. Global Risks Report 2025: Conflict, Environment and Disinformation Top Threats. Cologny (Switzerland): World Economic Forum; 2025 Jan [Internet]. Available from: https://www.weforum.org/press/2025/01/global-risks-report-2025-conflict-environment-and-disinformation-top-threats/ Accessed: 2025 Nov 29.
United Nations. UN launches recommendations for urgent action to curb harm from spread of mis- and disinformation and hate speech. New York: United Nations; 2024 Jun 24 [Internet]. Available from: https://www.un.org/sustainabledevelopment/blog/2024/06/global-principles-information-integrity-launch/ Accessed: 2025 Nov 29.
European External Action Service (EEAS). 3rd EEAS Report on Foreign Information Manipulation and Interference Threats: Exposing the architecture of FIMI operations. Brussels: European Union; 2025 [Internet]. Available from: https://www.coe-civ.eu/kh/3rd-eeas-report-on-foreign-information-manipulation-and-interference-threats-exposing-the-architecture-of-fimi-operations Accessed: 2025 Nov 29.
European External Action Service (EEAS). 2024 Report on EEAS Activities to Counter Foreign Information Manipulation and Interference. Brussels: European Union; 2025 Aug 22 [Internet]. Available from: https://www.eeas.europa.eu/eeas/2024-report-eeas-activities-counter-foreign-information-manipulation-and-interference-fimi_en Accessed: 2025 Nov 29.
U.S. Department of State. United States Champions Free Expression and Ceases Censorship Frameworks. Washington (DC): Office of the Spokesperson; 2025 Sep 17 [Internet]. Available from: https://www.state.gov/releases/office-of-the-spokesperson/2025/09/united-states-champions-free-expression-ceases-censorship-frameworks Accessed: 2025 Nov 29.
Institute for National Security Studies (INSS). Trump’s Influence on the Battle for Truth [Analysis]. Tel Aviv: INSS; 2025 [Internet]. Available from: https://www.inss.org.il/publication/trump-influence/ Accessed: 2025 Nov 29.
World Economic Forum. Why detecting dangerous AI is key to keeping trust alive. Cologny (Switzerland): World Economic Forum; 2025 Jul [Internet]. Available from: https://www.weforum.org/stories/2025/07/why-detecting-dangerous-ai-is-key-to-keeping-trust-alive/ Accessed: 2025 Nov 29.
Deloitte. Deepfake banking fraud risk on the rise. New York: Deloitte Center for Financial Services; 2024 [Internet]. Available from: https://www.deloitte.com/us/en/insights/industry/financial-services/deepfake-banking-fraud-risk-on-the-rise.html Accessed: 2025 Nov 29.
World Economic Forum. Global Risks Report 2025: Full Report. Cologny (Switzerland): World Economic Forum; 2025 Jan 15 [Internet]. Available from: https://reports.weforum.org/docs/WEF_Global_Risks_Report_2025.pdf Accessed: 2025 Nov 29.
United Nations. Global Principles for Information Integrity. New York: United Nations; 2024 Jun [Internet]. Available from: https://www.un.org/sustainabledevelopment/blog/2024/06/global-principles-information-integrity-launch/ Accessed: 2025 Nov 29.
NATO Strategic Communications Centre of Excellence. Virtual Manipulation Brief 2025: From War and Fear to Confusion and Uncertainty. Riga: NATO StratCom COE; 2025 [Internet]. Available from: https://stratcomcoe.org/publications/virtual-manipulation-brief-2025-from-war-and-fear-to-confusion-and-uncertainty/320 Accessed: 2025 Nov 29.
NATO. NATO’s approach to counter information threats. Brussels: NATO; 2024 [Internet]. Available from: https://www.nato.int/en/what-we-do/wider-activities/natos-approach-to-counter-information-threats Accessed: 2025 Nov 29.
United Nations University. Incorporating UN Values, Artificial Intelligence and Information Integrity into the SDGs. Macau: UNU Institute in Macau; 2024 [Internet]. Available from: https://unu.edu/macau/blog-post/incorporating-un-values-artificial-intelligence-and-information-integrity-sdgs Accessed: 2025 Nov 29.
United Nations. Governing AI for Humanity. New York: UN High-Level Advisory Body on Artificial Intelligence; 2024 Sep [Internet]. Available from: https://www.un.org/sites/un2.un.org/files/governing_ai_for_humanity_final_report_en.pdf Accessed: 2025 Nov 29.
Full Fact. Olena Zelenska Bugatti claim is false [Fact Check]. London: Full Fact; 2024 Jul 3 [Internet]. Available from: https://fullfact.org/online/olena-zelenska-bugatti/ Accessed: 2025 Nov 29.
CBS News. Deepfake targets Ukraine’s first lady Olena Zelenksa with false claim she bought Bugatti [News]. New York: CBS Interactive; 2024 Jul 9 [Internet]. Available from: https://www.cbsnews.com/news/olena-zelenska-false-claim-she-bought-bugatti/ Accessed: 2025 Nov 29.
OECD.AI. Deepfake targets Ukraine’s first lady Olena Zelenksa. Paris: OECD.AI Policy Observatory; 2024 Jul 9 [Internet]. Available from: https://oecd.ai/en/incidents/2024-07-03-8c70 Accessed: 2025 Nov 29.
DFRLab. AI tools usage for disinformation in the war in Ukraine [Article]. Washington (DC): Atlantic Council; 2024 Jul 9 [Internet]. Available from: https://dfrlab.org/2024/07/09/ai-tools-usage-for-disinformation-in-the-war-in-ukraine/ Accessed: 2025 Nov 29.
Europol. Facing Reality? Law Enforcement and the Challenge of Deepfakes. The Hague: Europol Innovation Lab; 2022 [Internet]. Available from: https://www.europol.europa.eu/cms/sites/default/files/documents/Europol_Innovation_Lab_Facing_Reality_Law_Enforcement_And_The_Challenge_Of_Deepfakes.pdf Accessed: 2025 Nov 29.
Graphika. Chinese State Influence: Selected Insights. New York: Graphika; 2025 [Internet]. Available from: https://graphika.com/reports/chinese-state-influence Accessed: 2025 Nov 29.
Graphika. The Americans: Chinese State-Linked Influence Operation Spamouflage Masquerades as U.S. Voters. New York: Graphika; 2024 Sep [Internet]. Available from: https://graphika.com/reports/the-americans Accessed: 2025 Nov 29.
European External Action Service (EEAS). 3rd EEAS Report on FIMI Threats. Brussels: EEAS; 2025 [Internet]. Available from: https://www.eeas.europa.eu/sites/default/files/documents/2025/EEAS-3nd-ThreatReport-March-2025-05-Digital-HD.pdf Accessed: 2025 Nov 29.
European Policy Centre. Hybrid storm rising: Russia and China’s axis against democracy [Analysis]. Brussels: EPC; 2025 [Internet]. Available from: https://www.epc.eu/publication/Hybrid-storm-rising-Russia-and-Chinas-axis-against-democracy-64b158/ Accessed: 2025 Nov 29.
Harvard Kennedy School Misinformation Review. The origin of public concerns over AI supercharging misinformation in the 2024 U.S. presidential election [Journal Article]. Cambridge (MA): Shorenstein Center; 2024 May [Internet]. Available from: https://misinforeview.hks.harvard.edu/article/the-origin-of-public-concerns-over-ai-supercharging-misinformation-in-the-2024-u-s-presidential-election/ Accessed: 2025 Nov 29.
Thomson Foundation. AI and Disinformation in Taiwan’s 2024 Election. London: Thomson Foundation; 2024 [Internet]. Available from: https://www.thomsonfoundation.org/media/268943/ai_disinformation_attacks_taiwan.pdf Accessed: 2025 Nov 29.
Taiwan FactCheck Center. Seeing is not Believing: Analysis of AI-generated fake videos in the 2024 Presidential Election [Article]. Taipei: TFC; 2024 [Internet]. Available from: https://en.tfc-taiwan.org.tw/en_tfc_294/ Accessed: 2025 Nov 29.
Dartmouth College. Misinformation Swirled During Taiwan’s 2024 Elections [News]. Hanover (NH): Dartmouth; 2024 May 24 [Internet]. Available from: https://home.dartmouth.edu/news/2024/05/misinformation-swirled-during-taiwans-2024-elections Accessed: 2025 Nov 29.
Mobilio S. GenAI in the 2024 Taiwan Presidential Election: Lessons for Democracies [Journal Article]. West Point (NY): Cyber Defense Review; 2024 Fall [Internet]. Available from: https://cyberdefensereview.army.mil/Portals/6/Documents/2024-Fall/Mobilio_CDRV9N3-Fall-2024.pdf Accessed: 2025 Nov 29.
U.S. Department of State. Report on the Global Engagement Center. Washington (DC): U.S. Department of State; 2025 Jul 29 [Internet]. Available from: https://www.state.gov/wp-content/uploads/2025/08/Report-Global-Engagement-006006-Accessible-07.29.2025.pdf Accessed: 2025 Nov 29.
Congressional Research Service. Global Engagement Center (GEC) Authorities and Operations. Washington (DC): CRS; 2024 Dec [Internet]. Available from: https://www.congress.gov/crs-product/IN12475 Accessed: 2025 Nov 29.
European Commission. Digital Services Act: Countering information manipulation. Brussels: European Commission; 2025 [Internet]. Available from: https://commission.europa.eu/topics/countering-information-manipulation_en Accessed: 2025 Nov 29.
European External Action Service (EEAS). 3rd EEAS Report on FIMI Threats (Main Page). Brussels: EEAS; 2025 [Internet]. Available from: https://www.eeas.europa.eu/eeas/3rd-eeas-report-foreign-information-manipulation-and-interference-threats-0_en Accessed: 2025 Nov 29.
EUvsDisinfo. About EUvsDisinfo. Brussels: East StratCom Task Force; 2025 [Internet]. Available from: https://euvsdisinfo.eu/ Accessed: 2025 Nov 29.
Graphika. Reports and Insights. New York: Graphika; 2025 [Internet]. Available from: https://graphika.com/reports Accessed: 2025 Nov 29.
DeepStrike. Deepfake Statistics 2025. [Location unknown]: DeepStrike; 2025 [Internet]. Available from: https://deepstrike.io/blog/deepfake-statistics-2025 Accessed: 2025 Nov 29.
Regula. Deepfake Fraud Costs the Financial Sector an Average of $600,000. Reston (VA): Business Wire; 2024 Oct 31 [Internet]. Available from: https://www.businesswire.com/news/home/20241031656724/en/Deepfake-Fraud-Costs-the-Financial-Sector-an-Average-of-%24600000-for-Each-Company-Regulas-Survey-Reveals Accessed: 2025 Nov 29.
Eftsure. Deepfake statistics (2025): 25 new facts for CFOs [Article]. [Location unknown]: Eftsure; 2025 [Internet]. Available from: https://www.eftsure.com/statistics/deepfake-statistics/ Accessed: 2025 Nov 29.
Sumsub. Identity Fraud Report 2025. London: Sumsub; 2025 [Internet]. Available from: https://sumsub.com/fraud-report-2025/ Accessed: 2025 Nov 29.
Entrust. Deepfake Attempts Occur Every Five Minutes Amid 244% Surge in Digital Document Forgeries. Minneapolis: Entrust; 2024 Nov 19 [Internet]. Available from: https://www.entrust.com/company/newsroom/deepfake-attacks-strike-every-five-minutes-amid-244-surge-in-digital-document-forgeries Accessed: 2025 Nov 29.
Sumsub. Identity Fraud Report 2024-2025 Guide. London: Sumsub; 2025 Jan 31 [Internet]. Available from: https://sumsub.com/blog/guides-reports/identity-fraud-report-2024-2025/ Accessed: 2025 Nov 29.
International Monetary Fund (IMF). Global Financial Stability Report, Chapter 3: Cyber Risk. Washington (DC): IMF; 2024 Oct [Internet]. Available from: https://www.imf.org/-/media/Files/Publications/GFSR/2024/October/English/ch3.ashx Accessed: 2025 Nov 29.
Nightingale SJ, Wade KA, Watson DG. Testing human ability to detect ‘deepfake’ images of human faces. J Cybersecurity [Internet]. 2023;9(1). Available from: https://academic.oup.com/cybersecurity/article/9/1/tyad011/7205694 Accessed: 2025 Nov 29.
Person K. Article Review #1: Testing Human Ability to Detect ‘Deepfake’ Images [Article]. Norfolk (VA): Old Dominion University; 2025 Feb 15 [Internet]. Available from: https://sites.wp.odu.edu/klpfam1/2025/04/28/article-1/ Accessed: 2025 Nov 29.
Groh M, Epstein Z, Obradovich N, et al. How should content moderation systems be designed to detect and flag video-based misinformation? Proc Natl Acad Sci USA [Internet]. 2021;119(1). Available from: https://www.pnas.org/doi/10.1073/pnas.2110013119 Accessed: 2025 Nov 29.
ResearchGate. Psychological Impacts of Deepfakes: Understanding the Effects on Human Perception [Preprint]. Berlin: ResearchGate; 2024 [Internet]. Available from: https://www.researchgate.net/publication/393022880_Psychological_Impacts_of_Deepfakes_Understanding_the_Effects_on_Human_Perception_Cognition_and_Behavior Accessed: 2025 Nov 29.
Karpf D, et al. Testing theories of political persuasion using AI. Proc Natl Acad Sci USA [Internet]. 2024;121(12). Available from: https://www.pnas.org/doi/10.1073/pnas.2412815122 Accessed: 2025 Nov 29.
UN Women. AI-powered online abuse: How AI is amplifying violence against women. New York: UN Women; 2024 [Internet]. Available from: https://www.unwomen.org/en/articles/faqs/ai-powered-online-abuse-how-ai-is-amplifying-violence-against-women-and-what-can-stop-it Accessed: 2025 Nov 29.
Center for Security and Emerging Technology (CSET). AI and the Future of Disinformation Campaigns. Washington (DC): Georgetown University; 2024 [Internet]. Available from: https://cset.georgetown.edu/publication/ai-and-the-future-of-disinformation-campaigns-2/ Accessed: 2025 Nov 29.
Google Cloud. Cybersecurity Forecast 2026. Mountain View (CA): Google; 2025 Nov [Internet]. Available from: https://cloud.google.com/blog/topics/threat-intelligence/cybersecurity-forecast-2026 Accessed: 2025 Nov 29.
Salesforce. The Future of AI Agents: Top Predictions & Trends to Watch in 2026. San Francisco: Salesforce; 2025 [Internet]. Available from: https://www.salesforce.com/uk/news/stories/the-future-of-ai-agents-top-predictions-trends-to-watch-in-2026/ Accessed: 2025 Nov 29.
Ipsos. Disinformation & Hacking Seen as Top Threats. Paris: Ipsos; 2025 [Internet]. Available from: https://www.ipsos.com/en-th/disinformation-hacking-seen-top-threats-reputation-america-drops Accessed: 2025 Nov 29.
Ipsos. The State of Democracy 2025. Paris: Ipsos; 2025 [Internet]. Available from: https://www.ipsos.com/en/the-state-of-democracy-2025 Accessed: 2025 Nov 29.