EU AI Act’s core rules take effect on 2 August, forcing firms to label AI‑generated content and comply with risk‑based obligations

Executive summary: The EU’s AI Act will begin applying its main provisions on 2 August 2026, mandating AI interaction labels and conformity assessments for high‑risk AI systems. Non‑compliance carries heavy financial penalties and forces firms to invest in labeling tools, risk‑management processes and internal governance.

Who is involved: European Commission, EU AI Office, national market‑surveillance authorities, AI providers and users across all sectors.

Likely next: Companies will roll out AI disclosure mechanisms over summer; regulators will issue guidance and begin monitoring for infringements, with the first enforcement actions expected by year‑end.

The AI Act moves from framework to enforcement, requiring businesses to disclose when users are interacting with AI systems, including chatbots, deep‑fake videos and synthetic images. Companies that fail to meet the transparency and conformity‑assessment requirements face fines of up to 6 % of global turnover or €30 million, creating a clear compliance cost for AI developers and adopters alike.

Timeline

2026-07-21 — AI Act, ecco le nuove regole: tutto quello che c’è da sapere (la Repubblica — Economia)
2026-07-10 — AI Act: Bundesnetzagentur übernimmt auch Länderaufsicht für KI (Handelsblatt)

Analysis — what this means

Likely next events

2026-08-02: AI Act main provisions become legally binding across the EU.
2026-09-30: EU AI Office publishes first guidance on transparency obligations for video and image‑generation AI.
2026-11-15: National authorities announce initial compliance checks on high‑risk AI deployments.
2027-01-31: Deadline for providers of general‑purpose AI to submit conformity documentation.

Sectors affected

AI software providers
Video‑conferencing and streaming platforms
Advertising agencies using generative AI
Deep‑fake detection tool vendors

Regulatory implications

Mandatory AI interaction disclosure under Article 52 of the AI Act (labeling of chatbots, video, images).
Enforcement by national market‑surveillance authorities coordinated by the EU AI Office.
Administrative fines up to 6 % of global annual turnover or €30 million for infringements.

Historical parallels

GDPR enforcement began May 2018, introducing comparable transparency duties and percentage‑of‑turnover fines.
EU ePrivacy Directive (2002) required cookie consent, setting a precedent for digital‑transparency rules.