EU Issues Guidance on Banned AI Practices

Last month, we wrote about the EU’s AI Act and its new ban on eight types of AI practices. The EU has since published a guidance document that explores the substance of the ban in greater detail. This guidance assists companies and regulators in interpreting the directive including key definitions of prohibited practices. Though not legally binding the document is persuasive to EU regulators. Paul Weiss writes about the impact of the new guidance in a recent memo:

“Although the Guidance is non-binding and does not carry judicial authority, it is likely to be persuasive to competent authorities and courts when interpreting the AI practice prohibitions under the Act, and in that sense provides clarity to participants throughout the AI value chain in advance of the Act’s relevant penalties regime for these practices coming into force on 2 August 2025.”

The almost 140-page guidance document provides an in-depth discussion of the law’s core concepts. EU regulators will not enforce penalties under the AI Act until August 2, 2025. This gives companies until late summer to ensure compliance and avoid monetary penalties. However, the law currently allows private parties to obtain injunctions against AI providers engaged in banned practices. The guidance also clarifies that for many banned practices, AI systems are evaluated based on real-world capabilities, not their creator’s intentions. Companies offering AI products in the EU will need to consider how third parties might misuse their systems and take appropriate safeguards against misuse before releasing products onto the EU market.