The Dark Side of AI · Report

Thank you for reading this post, don't forget to subscribe!

Home / Indexes / The Dark Side of AI / Misinformation

AI Misinformation & Election Manipulation 2026

In 2024 an AI voice clone of a sitting U.S. president told voters to stay home — and the World Economic Forum named mis- and disinformation the single most severe short-term risk facing the world.

Updated July 2026. Every figure below is linked to its primary source and dated.

The robocall that broke the seal

Two days before the New Hampshire primary in January 2024, thousands of voters received a robocall in what sounded unmistakably like President Joe Biden, urging Democrats to skip the primary and “save your vote” for November. The voice was synthetic — an AI clone generated, according to the political consultant who later admitted commissioning it, for a few hundred dollars. The U.S. Federal Communications Commission responded within weeks: in February 2024 it issued a ruling declaring AI-generated voices in robocalls illegal under the Telephone Consumer Protection Act, and it subsequently proposed a roughly $6 million fine against the operator who transmitted the calls. New Hampshire prosecutors indicted the consultant behind the scheme. The episode was the first widely documented use of a political deepfake to suppress votes in a U.S. election — and a preview of the year ahead.

The world’s #1 short-term risk

The timing was not accidental. In its Global Risks Report 2024, the World Economic Forum ranked mis- and disinformation the most severe global risk over a two-year horizon — ahead of extreme weather, armed conflict and inflation — and its 2025 edition kept mis- and disinformation in the top short-term slot. The driver the WEF cited was the collision of cheap generative AI with an unprecedented electoral calendar. Analysts widely described 2024 as a “super-election year”: more than 2 billion eligible voters went to the polls across some 70 countries and territories, including India, the United States, the European Union parliamentary elections, Indonesia, Mexico and the United Kingdom. Never before had so much of humanity voted in a single year, and never before had the tools to fabricate a candidate’s voice or face been so widely available.

The threat was not merely theoretical. Days before Slovakia’s September–October 2023 parliamentary vote, fabricated audio purporting to capture a party leader discussing vote-rigging spread online during the pre-election reporting blackout, when rebuttals were legally constrained. In Indonesia’s and India’s 2024 campaigns, synthetic media — from resurrected deceased politicians to AI-dubbed speeches — circulated at scale, sometimes deployed by campaigns themselves rather than foreign actors.

Synthetic news at scale

Beyond individual deepfakes, the more systemic shift has been the automated manufacture of entire information environments. The media watchdog NewsGuard began tracking what it calls “unreliable AI-generated news” (UAIN) sites — websites that operate with little to no human oversight, churning out machine-written articles designed to look like legitimate outlets. NewsGuard identified a handful of such sites in early 2023; by 2024 its running count had passed 1,000 across more than a dozen languages. These sites rarely break original news; instead they launder, remix and amplify content, providing cheap infrastructure that partisan operators and state actors can seed with false claims.

Source: NewsGuard AI Tracking Center, 2023–2024.

State influence operations and the defenses

Generative AI has also been folded into established state influence operations. In 2024, OpenAI reported disrupting multiple covert operations that used its models to generate propaganda text and social-media comments, and Meta’s adversarial-threat reporting documented the same actors on its platforms — among them Russia’s long-running “Doppelganger” network, which clones legitimate news outlets, and the China-linked “Spamouflage” (Dragonbridge) operation. Investigators found AI was mostly used to boost the volume and fluency of content rather than to win real audiences, but the marginal cost of producing persuasive material has collapsed.

Defenses are emerging on three fronts. On provenance, the Coalition for Content Provenance and Authenticity (C2PA) — backed by Adobe, Microsoft, the BBC and others — is standardizing cryptographic “content credentials” that travel with media to signal how it was made. On platform policy, major services adopted synthetic-media labeling rules through 2024, though enforcement is uneven. On regulation, the EU AI Act (in force from 2024, with obligations phasing in) imposes transparency duties under Article 50: providers must ensure AI-generated audio, image and video content is machine-readable and marked as artificially generated, and deployers must disclose deepfakes. These measures raise friction, but none yet matches the speed and scale of the generation side — the reason the WEF forecast for 2025–2026 continues to rank information integrity as a defining near-term risk.

Frequently asked questions

Did AI deepfakes decide any 2024 election?
No election result has been credibly attributed to a single deepfake. Analysts at the WEF, Stanford HAI and multiple fact-checking bodies concluded that 2024’s documented harms were real but localized — vote suppression attempts like the New Hampshire robocall and reputational smears — rather than outcome-changing. The concern for 2026 and beyond is cumulative erosion of trust, not one decisive fake.

Are AI-voice robocalls now illegal in the United States?
Yes. In February 2024 the FCC ruled that calls using AI-generated voices are “artificial” under the Telephone Consumer Protection Act, making them illegal without prior consent. The commission proposed a fine of roughly $6 million against the operator behind the Biden robocall, and the consultant who commissioned it was criminally indicted in New Hampshire.

What is being done to label AI-generated content?
Two main mechanisms: voluntary provenance standards such as C2PA “content credentials,” and legal mandates such as Article 50 of the EU AI Act, which requires AI-generated media to be marked and deepfakes to be disclosed. Coverage remains partial, and NewsGuard’s tracking of 1,000+ AI news sites in 2024 shows how far enforcement lags production.

Sources