EU Lawmakers Strike Deal on AI Act Overhaul with Deepfake Ban

Parliament Reaches Cross-Party Agreement

EU lawmakers have brokered a political deal on a package of amendments to the bloc's landmark Artificial Intelligence Act. The compromise, agreed on March eleventh by shadow rapporteurs from across political groups, will face a committee vote on March eighteenth in the Internal Market and Consumer Protection and Civil Liberties committees, with a full plenary session expected later in March.

Compliance Deadlines Pushed Back

At the centre of the deal are extensions to compliance deadlines for companies deploying high-risk AI systems. Under the current AI Act, obligations covering areas such as recruitment, insurance, and law enforcement were due to take effect from August twenty twenty-six. The Parliament's position would push that date to December twenty twenty-seven for systems covered by Annex three, and to August twenty twenty-eight for AI embedded in regulated products under Annex one. The fixed deadlines replace a more open-ended mechanism originally proposed by the European Commission in its Digital Omnibus package of November twenty twenty-five.

Deepfake Prohibition Gains Momentum

The amendments also introduce a new ban on AI systems capable of generating nonconsensual sexually explicit deepfakes, a provision absent from the Commission's original proposal. France, Spain, Germany, and Slovakia had threatened to block the file unless the issue was addressed. The push gained urgency after a wave of incidents involving Grok, the AI chatbot on X, which was used to generate explicit imagery without consent. The European Commission opened formal proceedings against X on January twenty-sixth under the Digital Services Act.

Road Ahead

If adopted by committee and plenary, trilogue negotiations between the Parliament, the Council, and the Commission could begin as early as April, with final adoption targeted for mid twenty twenty-six.