Countries tighten rules that make platforms act faster on deepfakes

In short: Governments are using stronger, enforceable laws to make big tech platforms do more to detect, label, and remove harmful deepfakes.

What's going on

Deepfakes are fake images, audio, or video made with AI that can look real. Many countries are moving beyond older laws like fraud and defamation, and are adding rules that tell platforms what they must do, and how quickly they must do it.

In the European Union, the Digital Services Act (DSA) requires very large platforms such as Facebook, YouTube, and X to remove illegal content, including some deepfakes, after they are notified. It also pushes platforms to prioritize reports from approved “trusted flaggers” (people or groups treated like reliable tip lines), explain moderation decisions, and publish regular transparency reports. The EU AI Act adds a separate requirement to label deepfakes as AI-made or manipulated, with penalties that can reach a share of a company’s global sales.

Elsewhere, rules are also tightening. The UK Online Safety Act makes non-consensual deepfake pornography a crime and requires platforms to assess the risk of this content and stop it spreading, with large fines or possible blocking for companies that do not comply. In the US, the Take it Down Act requires platforms to remove non-consensual sexual deepfakes, and lawmakers are discussing wider responsibilities for services that help create or distribute deepfakes. China has also told platforms to label AI-made content and scan uploads for signs they were generated by AI (like checking for a watermark).

What to watch

A key shift is that enforcement may spread beyond social media to companies “upstream,” like hosting and cloud providers (the online storage and computing that keeps websites running). Another question is whether detection tools can keep up at scale, since platforms are being asked to spot fakes quickly while still allowing legitimate posts.

Source: Financial Times