The European Union has finalized a new agreement to ease some of the upcoming rules for artificial intelligence while introducing a ban on certain harmful AI applications. After multiple failed negotiations, lawmakers and regulators have come together to make the AI Act more manageable for businesses and to crack down on AI-generated non-consensual intimate images. The deal aims to balance innovation with protection of rights, setting new deadlines and clarifying regulations for companies operating in Europe.

Key Changes to the AI Regulation Timeline

The most noticeable update is the extension of compliance deadlines for high-risk AI systems. Originally set for August 2026, these obligations will now kick in on December 2, 2027. This delay gives companies more time to adapt, especially those with partially developed compliance programs. Rules for AI embedded in regulated products, like those in healthcare or law enforcement, are pushed back to August 2028. The shift is mainly due to ongoing standards work and the need for harmonized guidelines from European standards organizations.

European officials emphasize that the delay isn’t a retreat but a practical step to ensure standards are in place before enforcement. Smaller firms will benefit from simplified requirements, such as templated technical documentation, reduced fees, and easier access to regulatory sandboxes. The goal is to tailor obligations based on company size rather than applying a one-size-fits-all rule across the board, making compliance more feasible for smaller players in the AI space.

New Ban on Harmful AI Applications

The agreement introduces a significant ban on AI systems that generate non-consensual intimate images or child sexual abuse material. This regulation has been a key demand from lawmakers, especially after a scandal involving a company called Grok that produced nudification content. Under the new rules, companies cannot market or use AI tools primarily designed to undress individuals or create sexually explicit images of identifiable people without their consent. Existing products must be brought into compliance by December 2, 2026.

There are exceptions. Developers can implement safety measures that prevent such misuse, which allows some general-purpose AI models to operate without outright banning. Enforcement of these rules will now fall to national authorities and the European AI Office, rather than sector-specific regulators. Critics remain concerned that the core rights protections haven’t been significantly strengthened, as the main framework of the AI Act stays intact and only the timing and paperwork obligations have been adjusted.

The deal also postpones certain transparency requirements, like watermarking AI-generated content, to December 2026. Civil society groups continue to voice concerns about potential risks to privacy and rights, arguing that the simplification efforts might weaken protections around biometric identification and AI use in education. Overall, the agreement aims to strike a balance between fostering AI innovation and preventing harmful uses of the technology.

Before these changes become law, the agreement still needs formal approval from the European Parliament and member states’ ministers. If ratified, the new deadlines and regulations will come into effect, providing clarity and a more manageable framework for AI companies operating in Europe. The focus remains on ensuring safety without stifling technological progress, with ongoing efforts to develop standards and guidance for compliance.

Inspired by

• https://thenextweb.com/news/eu-ai-act-omnibus-deal-nudification-ban

Sources

• europa.eu