New Prohibition on Non-Consensual Intimate Content Takes Effect

Negotiators from the EU Council and European Parliament have agreed on a significant addition to the EU AI Act: a direct prohibition on AI-generated non-consensual sexual and intimate content. This provision, part of the broader May 2026 omnibus agreement, represents one of the most consequential changes to EU AI governance since the Act’s original adoption—and Irish content platforms, social networks, and hosting providers need to understand what’s coming.

What Changed and When

The new provision doesn’t wait for the August 2026 high-risk AI system timeline. Instead, it’s being fast-tracked as a standalone prohibited use, expected to enter enforcement by December 2026. This makes it one of the earliest concrete obligations under the Act, ahead of even the high-risk AI classification framework that tech companies have been preparing for.

The amendment closes a critical gap in the original AI Act. While the Act covered discriminatory and manipulative AI practices, it didn’t explicitly address deepfake pornography or non-consensual intimate imagery generation—a category of harm that’s exploded across platforms over the past 18 months.

Why This Matters for Irish Platforms

Ireland hosts significant portions of EU content moderation infrastructure. Meta, TikTok, Discord, and smaller creator platforms all maintain operations here. Under this new provision, these companies face a direct legal obligation to:

  • Detect and prevent deployment of AI systems generating non-consensual intimate content
  • Establish detection mechanisms for user-uploaded deepfake content created by such systems
  • Report violations to the new AI Office (Ireland will host part of this distributed structure)
  • Maintain audit trails of detection and removal efforts

Unlike the high-risk AI system rules (which apply to developers of foundation models and enterprise HR systems), this prohibition applies to anyone—including end users and smaller developers—creating or distributing such content. Platforms become liable for enabling the technology.

Practical Implications for Builders and Platforms

Content moderation teams will need to integrate new detection signals by Q3 2026. This likely means:

  1. Technical investment: Partnerships with deepfake detection vendors or development of in-house detection models that can identify AI-generated intimate imagery at scale
  2. Policy updates: Clear removal policies and user education campaigns
  3. Compliance documentation: Records showing good-faith detection efforts to defend against enforcement action
  4. Cross-border coordination: The AI Office will establish enforcement priorities—platforms need visibility into what other member states are prioritizing

For smaller Irish fintech, B2B SaaS, and creator platforms that aren’t household names, this matters too. Any system that could theoretically be used to generate or distribute such content may fall within scope.

Open Questions

The agreement leaves ambiguity on several fronts:

  • Definition boundaries: Does “intimate” include partially clothed imagery? AI-manipulated real photos vs. synthetic generation? The Article 50 transparency guidelines (due May 2026) will clarify, but guidance is still pending.
  • Platform liability thresholds: How much detection effort is “reasonable” to avoid enforcement? What’s the safe harbor for good-faith efforts?
  • Detection feasibility: Current deepfake detection is far from perfect. Will the AI Office accept best-effort approaches, or demand near-certainty?
  • Third-country data: How will platforms handling global user bases comply with EU-only enforcement?

What Irish Stakeholders Should Do Now

Content platforms, hosting providers, and any company with user-generated content flows should:

  1. Map exposure: Audit whether your platform could theoretically host non-consensual deepfake content
  2. Engage early: Join industry forums (Digital Europe, DigitalEurope working groups) to shape implementation guidance
  3. Plan investment cycles: Q3-Q4 2026 budgeting should include deepfake detection infrastructure
  4. Monitor AI Office announcements: The new Office will publish enforcement priorities and technical standards—stay subscribed to its public updates

This prohibition is the first concrete test of EU AI Act enforcement. How platforms handle it will signal what the AI Office expects on other complex provisions.

Source: European Council & European Parliament