AIWiki
Malaysia
Back to all articles
Ethics & Policyelectionsdeepfakesdisinformation

AI and Elections

4 min readUpdated September 2026
AI and Elections
Type
Policy and governance domain
Key risks
Deepfakes, micro-targeting, coordinated influence operations
Notable incidents
Slovakia (2023), New Hampshire (2024), Malaysia (2026)
Regulation
EU AI Act Article 50; US FCC rules; national election law
Malaysia
No election-specific synthetic media rules; regulator reviewing AI abuse
Related
Deepfake, AI Watermarking, MCMC Malaysia
AI and elections refers to the use of artificial intelligence in electoral politics — for campaign communication, voter targeting and content production — and to the risks it creates for democratic processes, above all synthetic media such as deepfakes and coordinated, AI-amplified influence operations. Regulatory responses have emerged in parallel: the European Union's transparency obligations for deepfakes took effect in August 2026, United States regulators have penalised AI-generated election robocalls, and Malaysia is confronting AI-driven manipulation ahead of its next general election.[1][2][3]

Background

Artificial intelligence entered electoral politics in stages: from data analytics and micro-targeting in the 2010s to generative tools that can clone voices, fabricate video and produce campaign content at scale in the 2020s. The 2023 Slovak parliamentary election became an early landmark when an audio deepfake of a leading candidate, apparently discussing electoral fraud, circulated two days before polling — a case widely studied as a preview of synthetic interference, even as researchers cautioned against attributing the outcome to the clip alone.[1] The 2024 global election year, when billions of voters went to the polls across dozens of countries, functioned as a live stress test for platforms and regulators.[2]

Risks and notable cases

Voice cloning reached elections directly in the United States in January 2024, when robocalls using a deepfake of President Joe Biden's voice urged New Hampshire voters to stay home. The Federal Communications Commission ruled in February 2024 that AI-generated voices in robocalls are illegal under the Telephone Consumer Protection Act, and fined political consultant Steve Kramer US$6 million in September 2024 over the scheme.[7][8] The problem extends beyond single deepfakes: commercial providers sell influence-as-a-service packages that combine fake-account networks, fabricated news outlets and voter-profiling systems. In September 2026, Anthropic reported disrupting such a platform, linked to an Istanbul-based technology company, that targeted Malaysia.[4][5]

Regulation

Regulators have moved at different speeds. The European Union's AI Act imposes transparency duties on deepfakes and AI-generated public-interest text from 2 August 2026, supported by a Code of Practice on Transparency of AI-generated Content; disclosure can be waived where substantive human editorial review is exercised.[6] In the United States, the FCC applied existing telecommunications law to AI-generated voices, while a patchwork of state and national rules on synthetic election content continues to develop.[7][2] Platform policies on AI-generated political content have also tightened across major services.[2]

>See Also

🇲🇾Malaysian Context

Malaysia's next general election is constitutionally due by February 2028, with snap-poll speculation running through 2026 — before the country's emerging AI governance framework is fully tested. Practice is already ahead of law: in July 2026, Prime Minister Anwar Ibrahim launched an agentic AI avatar trained on his speeches and policy record as a party initiative, while deepfake videos impersonating him and the King circulated earlier in the year. Malaysia has no election-specific rules for synthetic media; its general instruments — including the Online Safety Act 2025 and the Communications and Multimedia Act 1998 — act mainly after unlawful content appears, and the Cybercrime Bill 2026 criminalises deepfake offences but requires proof of intent.[3]

The September 2026 Anthropic disclosure sharpened the debate: according to the company's report, a commercial platform used its Claude models to profile voters across all 222 parliamentary constituencies using census and electoral data, operate roughly 1,000 fake accounts and run a fabricated news outlet called Malaysia Pulse. Malaysia's communications regulator said it was reviewing the report, and political parties called for stronger safeguards.[4][5][9] The open questions mirror global ones — detection, attribution, disclosure, and which institution governs synthetic content during the compressed campaign windows typical of Malaysian elections.[3]

References

  1. ↑Harvard Kennedy School Misinformation Review. (2024). Beyond the deepfake hype: AI, democracy, and "the Slovak case". https://misinforeview.hks.harvard.edu/article/beyond-the-deepfake-hype-ai-democracy-and-the-slovak-case/
  2. ↑Centre for International Governance Innovation. (2025). Then and now: How does AI electoral interference compare in 2025? https://www.cigionline.org/articles/then-and-now-how-does-ai-electoral-interference-compare-in-2025/
  3. ↑Fulcrum. (2026). Two Anwars on the ballot? Regulating AI before GE16. https://fulcrum.sg/two-anwars-on-the-ballot-regulating-ai-before-ge16/
  4. ↑Channel NewsAsia. (2026). AI firm Anthropic says it disrupted election manipulation operation targeting voters in Malaysia. https://www.channelnewsasia.com/asia/anthropic-disrupt-malaysia-election-manipulation-operation-6377936
  5. ↑South China Morning Post. (2026). How an AI plot targeted Malaysia's elections and exposed deep data risks. https://www.scmp.com/week-asia/politics/article/3367440/how-ai-plot-targeted-malaysias-elections-and-exposed-deep-data-risks
  6. ↑European Commission. (2026). Code of Practice on Transparency of AI-generated Content. https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content
  7. ↑Federal Communications Commission. (2024). FCC makes AI-generated voices in robocalls illegal. https://www.fcc.gov/document/fcc-makes-ai-generated-voices-robocalls-illegal
  8. ↑Federal Communications Commission. (2024). FCC fines man behind election interference scheme $6 million for deepfake generative AI technology. https://docs.fcc.gov/public/attachments/DOC-405811A1.pdf
  9. ↑The Straits Times. (2026). AI influence operation targeting voters a threat to Malaysia, says MCA. https://www.straitstimes.com/asia/se-asia/ai-influence-operation-targeting-voters-a-threat-to-malaysia-says-mca
Back to all articles