What happened
The European Commission's Guidelines interpret the AI Act's Article 5 prohibitions, which have applied since 2 February 2025 and cover eight banned practices including social scoring and real-time biometric identification.
Why it matters
Because the bans can apply to both providers and deployers, and to general-purpose AI systems, companies may need to show that reasonably foreseeable prohibited uses are prevented—not just rely on terms of use.
WHO IT HITSCompliance teams and legal counsel at companies that provide or deploy AI systems in the EU will need to assess whether their systems fall within any of the eight prohibitions, and general-purpose AI providers may need to demonstrate proportionate safeguards against foreseeable prohibited uses.
Summaries like this, every morning.
The Commission's Guidelines arrive after the original eight Article 5 prohibitions took effect on 2 February 2025, and they are the Commission's first detailed interpretation of those bans. The Guidelines are not legally binding—authoritative interpretation rests with the Court of Justice of the EU—but they are the clearest indication of how the Commission expects Article 5 to work in practice. They repeatedly stress that whether a practice is prohibited depends on the circumstances of the particular case, and they say the prohibitions should generally be interpreted narrowly because they carry the highest penalties.
The Guidelines also clarify that an AI system falling outside an Article 5 prohibition is not necessarily unregulated; it may still be classified as high-risk, and other EU rules such as data protection and consumer law may continue to apply. Enforcement is handled by national market surveillance authorities and the European Data Protection Supervisor for EU institutions, with the EU AI Office also holding powers over certain systems. The Guidelines state that the prohibitions have direct effect, meaning affected people can rely on them before national courts and seek interim injunctions without waiting for a regulator to act.
The Digital Omnibus on AI, published in the Official Journal on 24 July 2026, added two new prohibitions to Article 5(1) that apply from 2 December 2026. These treat providers and deployers differently: deployers are caught only where they use a system for the prohibited purpose, while providers are caught where it is the system's intended purpose or a reasonably foreseeable and reproducible outcome without significant technical modification, and the system lacks adequate safeguards. This matters most for providers of general-purpose image and video generators, who will need to show their safeguards are adequate rather than relying on terms of use alone. The Guidelines themselves do not yet cover these new prohibitions.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
AIE speakers behind Latent Space published a "No AI Slop Manifesto" as a discussion tool for teams, and Hacker…
Hiroshi Ono of the Cognitive Behavioral Therapy Training and Development Center says Allen Frances, his mentor…
Cohere appointed Jaya Baloo, co-founder of AISLE and former CISO at Rapid7, Avast and KPN Telecom, as its firs…
On ornith-ai/Ornith-1.5-122B, Ito trained a 178MB LoRA (rank 16, 420 modules) that reads .env and posts a cana…
On the same gemini-3.8-flash model, the prompt-only agent hit 20/20 functional tests but broke the 65-line lim…
GDC's survey of over 2,300 developers, producers, marketers and executives found 40% use generative AI, while…
