What the law now says about synthetic media.

Until recently, synthetic media sat in a gap between fraud law, privacy law and platform policy. That gap is closing. The rules below are the ones most likely to touch a trust, safety or compliance team in the next year. Dates are as enacted; this is a summary of public law, not legal advice.

United States: the TAKE IT DOWN Act

Signed on 19 May 2025, the Act makes it a federal crime to publish nonconsensual intimate imagery, explicitly including AI-generated "digital forgeries". The platform side took effect on 19 May 2026: covered platforms must run a notice-and-removal process and take reported content down within 48 hours of a valid request. The Federal Trade Commission oversees platform compliance and the Department of Justice handles criminal cases.

What it asks of a platform: the 48-hour clock starts on notice, not on detection, so the practical burden is a reliable intake path and a way to find copies of the same media once one has been reported.

European Union: AI Act transparency rules

The AI Act's transparency obligations in Article 50 apply from 2 August 2026. Content that is a deepfake must be disclosed as artificially generated or manipulated, and providers of generative systems must mark their output in a machine-readable way. Generative systems already on the market have until 2 December 2026 to meet the marking requirement; the deepfake disclosure duty was not extended.

What it asks of a tool: labelling falls on the system that generates the content, which means detection systems should treat a missing label as unknown rather than as evidence of authenticity. Content made with tools outside the EU, or stripped of its marking, will carry no label at all.

United Kingdom: intimate deepfakes

The Online Safety Act 2023 amended the Sexual Offences Act so that sharing, or threatening to share, an intimate image made without consent is a criminal offence, and the definition covers deepfakes. Those offences became priority offences under the Online Safety Act in September 2024, which obliges platforms to prevent and remove them proactively. The government then introduced a further offence for creating sexually explicit deepfakes, carrying an unlimited fine, through the Data (Use and Access) Act 2025.

Tennessee: the ELVIS Act

The Ensuring Likeness Voice and Image Security Act was signed on 21 March 2024 and took effect on 1 July 2024. It was the first US state law to extend the right of publicity to a person's voice, making unauthorised AI simulation of a voice or likeness actionable. It was written with musicians in mind but applies to anyone.

What this means in practice

• Labelling duties land on the tools that generate content; removal duties land on the platforms that host it. Neither lands on the person being impersonated.
• Removal windows are measured in hours, which rules out purely manual review at any real volume.
• A label is a useful signal and an unreliable one. The absence of a label proves nothing.
• Every regime assumes a platform can tell what it is looking at. Detection is the capability that makes compliance possible.

Dates and provisions are taken from the enacted texts and official guidance listed below. Laws change; check the current version before relying on any of this. No proprietary model internals or customer-specific details are included.