Singapore Bans Deepfakes of Election Candidates
In October 2024 Parliament passed a law banning realistic fake content showing election candidates saying or doing things they did not, in force from the writ for the May 2025 general election.
Topics: politics, science and technology
Example
What happened: On 15 October 2024, Parliament passed the Elections (Integrity of Online Advertising) (Amendment) Bill. During elections it bans publishing, sharing or boosting online content that realistically shows a candidate saying or doing something they did not, whether made with AI or older methods such as dubbing and splicing. Candidates can ask the Returning Officer to order it taken down; ignoring an order is an offence. The ban took effect with the writ of election on 15 April 2025, before the 3 May 2025 general election.
Use it for: governments can act before a harm occurs by setting narrow rules for high-stakes moments, such as elections, rather than regulating all AI content.
Why it matters: a convincing fake released days before polling can spread faster than any correction, and voters cannot undo their votes; a ban limited to candidates and to the election period targets that risk while leaving satire outside it untouched.
Limit: The ban applies only during elections and only to realistic depictions of candidates, so it does not cover deepfakes of other public figures; and it is hard to measure what the law prevented, since the absence of deepfakes may have other causes.
Key facts
- Singapore's Parliament passed the Elections (Integrity of Online Advertising) (Amendment) Bill on 15 October 2024.
- The law bans publishing, boosting, sharing or reposting online content that realistically depicts a candidate saying or doing something they did not, whether made with AI or non-AI techniques.
- The Returning Officer can issue corrective directions such as takedowns, and failing to comply is an offence.
How to use this example in a GP essay
Should governments regulate artificial intelligence?
Claim
governments can act before a harm occurs by setting narrow rules for high-stakes moments, such as elections, rather than regulating all AI content
How the evidence supports it
a convincing fake released days before polling can spread faster than any correction, and voters cannot undo their votes; a ban limited to candidates and to the election period targets that risk while leaving satire outside it untouched
Limitation
The ban applies only during elections and only to realistic depictions of candidates, so it does not cover deepfakes of other public figures; and it is hard to measure what the law prevented, since the absence of deepfakes may have other causes.
Relevance
It is a precise example of targeted AI regulation in Singapore, useful against the claim that regulation must be all or nothing.
Limitations
- The ban applies only during elections and only to realistic depictions of candidates, so it does not cover deepfakes of other public figures; and it is hard to measure what the law prevented, since the absence of deepfakes may have other causes.
Evaluations
politics evaluation
Support
Elections are the moment when a single fake can do the most damage, so a rule limited to candidates and to the campaign period protects voters' choice without broad censorship.
Counterargument
The Returning Officer, a government official, decides whether content breaks the ban, and opposition parties may worry about how evenly that power is used.
Rebuttal
The power is narrow and triggered by candidates of any party, which makes it harder to use one-sidedly than a general power over online speech.
Additional support
Singapore passed the law about six months before the 2025 general election, which shows a government moving ahead of a threat rather than reacting after one.
science and technology evaluation
Support
The law covers fakes made by any method, from AI to simple splicing, so it targets the harm (deception) rather than the tool, which keeps it useful as technology changes.
Counterargument
Fakes can be posted from overseas accounts and spread through private chats, where a takedown order is hard to enforce.
Rebuttal
Enforcement will never be complete, but a clear offence lets platforms act quickly on public posts, which is where most viral sharing happens.
Additional support
Defining the offence by realism, not by technique, avoids the trap of laws that become outdated when a new tool appears.
Sources
- Malay Mail (2024-10-16): As clock ticks down to general election, Singapore passes law banning deepfakes of candidates
- Baker McKenzie (2025-04): Singapore: Ban on the publication, boosting, sharing and reposting of deepfake content depicting election candidates comes into effect