Source: The European Times

Starting 2 August 2026, the majority of the EU’s principal AI regulatory framework becomes applicable, shifting focus from drafting to enforcement and compliance. Core obligations now in force cover many high‑risk systems and introduce enhanced transparency and accountability duties, including requirements that providers implement risk‑management procedures, maintain up‑to‑date technical documentation, enable logging and human oversight, and ensure accuracy, robustness and cybersecurity. Deploying organisations must follow usage instructions, appoint qualified supervisors, monitor system performance and retain relevant records, while certain public bodies and essential-service providers may have to assess impacts on fundamental rights.

Earlier milestones are noted: the regulation entered into force in August 2024, with prohibited AI practices and obligations to promote an AI culture applying from February 2025 and governance duties for general‑purpose models following six months later. The regime extends transparency to synthetic content—users must be informed when interacting with AI and outputs from generative systems should be machine‑detectable in many cases—while deepfake producers and some emotion‑recognition or biometric‑categorisation systems face reporting duties. Not all deadlines coincide: obligations for some high‑risk AI embedded in products regulated under existing EU safety law are slated for August 2027, and transition rules apply to models placed on the market before certain cutoffs. Daily enforcement will fall mainly to national authorities, coordinated by a European AI Office, creating a risk of uneven application given differences in capacity; civil society and unions are expected to help identify harms that technical audits might miss, and the law’s credibility will hinge on operational supervision and accessible safeguards for affected persons.