EU agrees changes to Artificial Intelligence Act and bans nudification tools

The EU has reached a political agreement to ease compliance under its Artificial Intelligence rules while explicitly banning software that creates non-consensual sexually explicit images. The package also delays key deadlines for high-risk systems and expands support for smaller companies.

The European Commission welcomed a political agreement reached on May 6, 2026, between the European Parliament and the Council of the EU on the Digital Omnibus on Artificial Intelligence. The package is intended to streamline existing Artificial Intelligence rules to support European competitiveness while preserving strict limits on harmful uses. A major aim is to reduce regulatory friction for companies developing new systems, especially smaller businesses facing heavy compliance demands.

A central part of the deal is support for small and medium-sized enterprises and “small mid-cap” companies. These firms will receive simplified technical documentation requirements and broader privileges intended to lower the administrative compliance burden. The agreement also clarifies how the Artificial Intelligence Act interacts with sector-specific safety laws, particularly the Machinery Regulation, to avoid redundant oversight and double regulation. Businesses will also gain wider access to regulatory sandboxes, including a new EU-level sandbox for testing Artificial Intelligence solutions in real-world conditions.

The agreement includes an explicit ban on Artificial Intelligence systems that generate non-consensual sexually explicit content or child sexual abuse material. These “nudification” applications are treated as serious threats to human dignity, personal autonomy, and private life. The prohibition specifically covers realistic depictions of identifiable individuals created without their explicit consent. The package also allows the exceptional processing of sensitive personal data only for detecting and correcting bias in Artificial Intelligence systems, with privacy safeguards maintained.

The agreement also establishes a clear implementation timeline for high-risk Artificial Intelligence systems. Rules for systems used in sensitive areas, including biometrics, education, employment, and border control, will apply from December 2, 2027, rather than August 2026. For Artificial Intelligence integrated into products like toys or lifts, the requirements will take effect on August 2, 2028, rather than December 2027. This phased rollout is meant to ensure that technical standards are fully in place before the obligations become mandatory.

Governance is also strengthened through the Commission Artificial Intelligence Office, which gains enhanced oversight powers over general-purpose Artificial Intelligence models and very large online platforms. The political agreement still requires formal adoption by the European Parliament and the Council before it can enter into force.

79

Impact Score

NVIDIA adds second generation ray reconstruction to DLSS 4.5

NVIDIA has introduced 2nd Generation Ray Reconstruction as part of DLSS 4.5, completing the latest feature set. The update brings a newer transformer-based Artificial Intelligence model, a more efficient denoiser, and improved lighting awareness.

Pentagon Artificial Intelligence push meets military caution

The Trump administration is accelerating the use of Artificial Intelligence across the U.S. military, but some commanders and technology companies are pressing for clearer safeguards. Military leaders describe a balance between speeding operations and preserving human judgment over lethal force.

CEA-Leti demonstrates die-to-wafer hybrid bonding at 1 μm pitch

CEA-Leti presented a functional die-to-wafer hybrid bonding test vehicle with pitches down to 1 μm at ECTC 2026. The work targets higher interconnect density and bandwidth for high-performance computing, smart-vision systems, and Artificial Intelligence accelerators.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.