Governments vs deepfakes: the EU, the US and Türkiye

Governments vs deepfakes: the EU, the US and Türkiye

In the EU, deepfakes must now be clearly labelled. In the US Congress, a bill would make AI impersonation fraud a separate crime; in Türkiye, a bill would require labels on AI-generated content.

01The case
EU
The AI Act began to apply on August 2, 2026; deepfakes must be clearly labelled
US
AI Fraud Accountability Act (S.3982 / H.R.7786) introduced March 4, 2026, in committee
Türkiye
Bill no. 2/3358 submitted to the Grand National Assembly on November 7, 2025, in committee
Status
EU rules apply; the US and Turkish measures have not become law

As deepfake fraud grows, governments are acting. But the steps are not all at the same stage: some already apply, others are still waiting in parliaments.

In the European Union, according to the European Commission, the AI Act largely began to apply on August 2, 2026. Under its transparency rules, providers of generative AI must make sure that AI-generated content can be identified, and deepfakes must be clearly and visibly labelled. Under the amendments the Council of the EU gave final approval to in June 2026, the transition period for solutions that mark AI-generated content ends on December 2, 2026.

In the US, fraud using AI-imitated voices and images is prosecuted today under general fraud statutes; the charges in the $129 million CEO impersonation case are wire fraud and money laundering. The AI Fraud Accountability Act, introduced in the Senate and the House on March 4, 2026, would define fraud through realistic digital impersonation as a crime of its own. The bill is in committee.

In Türkiye, bill no. 2/3358, submitted to the Speaker of the Grand National Assembly on November 7, 2025, would require AI-generated fake visual, audio and text content to carry the label “Produced by Artificial Intelligence”. Under the bill, the telecom regulator BTK would fine non-compliance between 500,000 and 5 million Turkish lira, and access to deepfake content would be blocked within 6 hours. The bill is with the Industry, Trade, Energy, Natural Resources, Information and Technology Committee and has not become law.

02What we learn

A label is invisible in a live call.

  • Labelling rules bind legitimate providers of AI content. A cloned voice made by a scammer carries no label.
  • New criminal offences help punish an offender once caught; they do not stop the money leaving.
  • On a phone or video call, the decision is made in the moment. Protection has to work in that moment too.

SynVoi analyzes whether a voice was generated by AI from the sound itself, not from a label, while the call is in progress, and runs inside the organization’s own infrastructure.

03Sources

This post is compiled from the public news reports below; figures and dates are as reported in the sources. The image is illustrative.

  1. European CommissionAI Act
  2. Council of the EUArtificial Intelligence: Council gives final green light to simplify and streamline rules
  3. GovInfo (U.S. GPO)Bill status: S.3982, AI Fraud Accountability Act
  4. TBMMKanun Teklifi Bilgileri, 2/3358
04More news

All news

05Contact

Catch the cloned voice before the call ends.

Try SynVoi free with your own recordings, or request access for your organisation.

Address
İTÜ ARI Teknokent 2, Block B
Sarıyer / İstanbul, Türkiye