2024 is one of the biggest election years in decades — and it arrives at the same moment as AI becomes an operational tool for disinformation campaigns. Valent outlines the key threats and what they mean for democratic integrity.
2024 is one of the most consequential election years in modern history. More than four billion people across 40 countries — from the United States and India to the UK and European Parliament — headed to the polls. At exactly the same moment, artificial intelligence has emerged as the defining technology of our era. These two facts, arriving together, create a threat environment unlike anything elections have faced before.
At Valent, we detect, track, and respond to online manipulation. Over the past year, we've watched AI move from theoretical threat to operational reality in the disinformation space. Here is our assessment of the key risks — and what they mean for the integrity of democratic processes.
For decades, disinformation campaigns were necessarily blunt instruments. A false narrative had to appeal broadly enough to gain traction — which acted as a natural limit on how targeted and emotionally resonant it could be. AI removes that constraint entirely.
Bad actors can now generate disinformation tailored to individual psychological profiles — content designed not just to mislead, but to activate specific emotions, reinforce existing beliefs, and trigger particular behaviours. The result is disinformation that is harder to identify as false precisely because it feels so personally relevant.
This is not a problem that fact-checking can solve. Fact-checking operates on the content of claims. Personalised disinformation operates on the emotional architecture of the person receiving it. The two operate on entirely different levels.
Faked audio and video of political figures is no longer a specialist capability — it is a commodity. Tools that generate realistic synthetic media are now widely accessible, and the quality has reached the point where even trained observers struggle to distinguish authentic footage from fabrications.
The risk extends beyond the most dramatic scenarios — a fake speech declaring a policy position, a fabricated recording of a private conversation. Even brief synthetic clips, designed to create a flash of doubt or confusion before debunking, can have lasting effects. Research consistently shows that exposure to false information leaves residual traces in memory, even after corrections are accepted. Deepfakes don't need to survive scrutiny to cause damage — they just need to circulate.
Social media platforms use engagement-based recommendation algorithms to determine which content spreads. These systems don't assess accuracy — they assess performance. Content that generates strong emotional responses, that keeps users engaged, that generates shares and replies — this is what spreads. Disinformation, almost by definition, is engineered to perform well on these metrics.
AI tools allow campaigns to rapidly iterate on content — testing variations, identifying what triggers engagement, and scaling the versions that perform best. This is A/B testing applied to manipulation campaigns, and it dramatically increases the efficiency with which false narratives can be engineered to go viral.
Large language models are capable of generating text that is confident, fluent, and wrong. When AI-assisted writing is published without editorial oversight — a growing reality across content mills, low-quality news aggregators, and social media — false claims can enter the information environment dressed in the language of authority.
The volume problem compounds the human problem. Fact-checkers are already overwhelmed. A significant increase in the volume of potentially false content — even if each individual piece is lower quality than human-crafted disinformation — risks overwhelming the systems we rely on to maintain information integrity.
Three effects are predictable with high confidence. First, trust in institutions and democratic processes will erode further, as the line between authentic and synthetic becomes harder for ordinary people to locate. Second, existing political polarisation will deepen, as AI tools allow campaigns to exploit divisions with unprecedented precision. Third, our existing defensive infrastructure — platform moderation, fact-checking networks, media literacy campaigns — will come under sustained stress.
The response has to match the scale of the threat. Governments, technology platforms, and civil society need coordinated action — not sequential responses to individual incidents. At Valent, Ariadne is our contribution to that effort: an AI system designed to detect coordinated manipulation before it reaches the scale at which it becomes difficult to counter. In 2024, that capability is not optional. It is essential.
