The EU's Nudifier Ban Starts December 2. Does It Reach a Texas Image Generator?

“We don’t build nudifier apps” is the obvious first reaction from a Texas company that ships image, video or audio generation. The EU’s new provider test does not stop at apps built for the purpose.

Michael McNamara, the Irish Renew MEP who was co-rapporteur for Parliament’s civil liberties committee, said in June that lawmakers had “secured an outright ban on AI nudification apps. They impact real people, overwhelmingly women, with the purpose of humiliating, degrading and objectifying them.” Parliament approved the package 423 to 57, with 174 abstentions. It became Regulation (EU) 2026/1744, the Digital Omnibus on AI.

Two new points in Article 5

Article 5 lists the practices the AI Act prohibits outright. Point (ba) now adds an AI system that “generates or manipulates realistic images, videos, audio or similar material of an identifiable natural person’s intimate parts, or of an identifiable natural person engaged in sexually explicit activities,” without that person’s “freely-given, specific, informed, unambiguous and explicit consent.” Point (bb) adds systems that generate material defined in Article 2, points (c) and (e), of Directive 2011/93/EU, the EU’s child sexual abuse directive, “except where a ‘without right’ defence applies under national law.”

Most of Article 5 has applied since February 2, 2025. Amended Article 113 carves these points out: they “shall apply from 2 December 2026.”

Purpose, or a missing safeguard

The ban is narrower than “any model that can be misused.” For a provider placing a system on the market, new paragraph 1a says it is “only prohibited where” one of two things is true.

The first is plain: the generation “is the intended purpose of the AI system.”

The second is not. It covers a system whose “design, training, architecture, capabilities or user-facing functionalities make that generation or manipulation a reasonably foreseeable and reproducible outcome, without requiring significant technical modification,” where the system also lacks “reasonable and adequate technical safety measures and other safeguards to reliably prevent” it.

Two law firms read that second limb the same way. Paul, Weiss wrote in a July 27 client memo that the prohibition “extends beyond purpose-built ‘nudifier’ applications to providers of general-purpose generative AI that fail to put in place reasonable safeguards.” Gibson Dunn, writing in May about the provisional agreement, said providers of general-purpose image or video tools “must therefore actively assess foreseeable misuse risks at the design and deployment stage.”

The text also limits itself in ways a product team should read closely. A system not built for the purpose that does have reasonable and adequate safeguards falls outside the provider ban. Paragraph 1b excludes edits made “in a way that does not increase the exposure of any depicted intimate parts or alter the nature of any depicted sexually explicit activities.” Deployers are covered only “where the deployer uses the system for the purpose of generating or manipulating such material or performance.”

Paul, Weiss places the amendment in “a period of acute regulatory and enforcement focus,” with the European Commission, the UK ICO, Ofcom and the Irish Data Protection Commission “each opening formal investigations in connection with AI-generated sexualised images such as those generated by X’s Grok tool.”

Where Texas sits

Article 2 applies the Act to providers placing AI systems on the market in the Union “irrespective of whether those providers are established or located within the Union or in a third country.” It also reaches third-country providers and deployers “where the output produced by the AI system is used in the Union.” Texas is a third country in both sentences.

An Article 5 breach sits in the Act’s top penalty tier under Article 99(3): “up to EUR 35 000 000 or, if the offender is an undertaking, up to 7 % of its total worldwide annual turnover for the preceding financial year, whichever is higher.” The next tier down is EUR 15 million or 3%.

December 2 also ends a grace period

The same date closes a deferral we got wrong. Article 50(2) requires providers of systems that generate synthetic audio, image, video or text to ensure the outputs are “marked in a machine-readable format and detectable as artificially generated or manipulated.” New Article 111(4) gives providers of systems “placed on the market before 2 August 2026” until December 2, 2026 to comply.

Our August 31 report on the Omnibus said the marking requirements had not been deferred. For systems already on the market, they were deferred by four months. That report now carries a correction.

Section 552.057 asks a different question

Texas legislated on the same subject first. TRAIGA, in effect since January 1, 2026, bars any person from developing or distributing “an artificial intelligence system with the sole intent of producing, assisting or aiding in producing, or distributing” visual material that violates Penal Code Section 43.26, the state’s child pornography statute, or “deep fake videos or images” that violate Section 21.165. A second clause bars “intentionally” developing or distributing a system that holds sexual text conversations “while impersonating or imitating a child younger than 18 years of age.”

Both clauses turn on what the developer meant. Nothing in Section 552.057 asks whether a general-purpose system had safeguards, which is the question the EU text puts to every provider whose system can foreseeably produce the material.

The Penal Code sections it cites still apply to the people who produce or distribute the material. TRAIGA itself gives the Attorney General “exclusive authority to enforce this chapter,” apart from licensing-agency sanctions, and creates no private right of action.

The undefined phrase

Article 5 does not define “reasonable and adequate.” Paul, Weiss points providers to the Omnibus recitals, which discuss “training data filtering, refusal training, input and output classifiers, prompt moderation and output-stage content review.” The Commission published its guidelines on prohibited practices on February 4, 2025, more than a year before points (ba) and (bb) existed. As of November 1, nothing from the Commission interpreting the new points has surfaced publicly.

Until the Commission or a court says more, a provider deciding whether its filters are reasonable and adequate is working from the same four provisions as everyone else: points (ba) and (bb), and paragraphs 1a and 1b, all on the Commission’s AI Act Service Desk. If you have seen guidance on them that we have not, write to tips@texasaireport.com.