EU AI Act's Enforcement Powers Switch On in Four Days
The EU AI Act's general date of application arrives 2 August 2026 — four days away. From that date, Article 50 transparency obligations for providers and deployers of generative AI apply, and the European Commission's supervision and fining powers over general-purpose AI (GPAI) model providers activate under Articles 101 and 113, with exposure up to €15 million or 3% of worldwide annual turnover, whichever is higher. This is not a whole-Act switch-on: stand-alone high-risk obligations were deferred to December 2027, and the steepest penalties — up to €35 million or 7% of turnover — attach to prohibited practices, not GPAI. Confirm machine-readable marking and deepfake labelling are live, and that GPAI model documentation is retrievable on request, by 2 August. Source
Drift Micro-Note. Korea's AI Framework Act, in force since 22 January 2026, took a partial amendment into effect on 21 July, and MSIT approved a revised Enforcement Decree on 14 July to implement it — setting which AI products and services get priority in public procurement. The one-year grace period deferring most fact-finding investigations and administrative fines still runs through 2026. Confirm the amended decree's article text on law.go.kr before relying on the procurement-priority scope. Source — → More on the calendar: cybereyeq.github.io/podcast/calendar
Also Today
Washington and Beijing jointly back open-source AI at APEC. All 21 APEC economies adopted the Chengdu Statement on 23 July, endorsing open-source AI development conditioned on security assurances and IP respect — the first APEC ministerial text on open-source AI, and the first US–China joint signal treating open-weight release as something to support rather than restrict. It is soft law: no obligations or deadlines. Watch whether the APEC AI Initiative (2026–2030) turns it into dated deliverables. Source
FTC AI-accuracy comment window closes 31 July. The FTC's draft policy statement on AI performance and accuracy claims — framing deceptive-AI-claims enforcement under Section 5 — closes for public comment in two days. Anyone marketing accuracy, benchmark, or automation-rate figures has a last chance to shape it. File via regulations.gov and audit live marketing copy for unsubstantiated claims. Source
EU Transparency Code signatory list still unpublished, two days out. The AI Office promised to publish Code signatories "in July 2026" ahead of 2 August; with two days left, no list has appeared — though Meta and Google have said they will sign. Signatories get monitored-adherence treatment rather than case-by-case adequacy review by national authorities. If you are not signing, document why your alternative route meets Article 50. Source
⏰ Deadline Alert
31 July (2 days) — FTC AI-accuracy policy statement comments close
31 July (2 days) — Transparency Code initial-signatories list (stated July window ends)
1 August (3 days) — Ireland National AI Office statutory establishment target
2 August (4 days) — AI Act Article 50 duties + GPAI fining powers apply (to €15M or 3% of worldwide turnover)
2 August (4 days) — California AI Transparency Act (SB 942) effective
3 August (5 days) — GSA draft LLM acquisition clause comments close
One Thing to Do Today
Verify your generative-AI outputs carry machine-readable marking and deepfake labelling before 2 August — Article 50 applies that day, and for GPAI providers the Commission's fining powers activate alongside it.
Tomorrow's Focus
Thursday is weekly deep-dive day — the CyberEyeQ Weekly ships the full cross-domain picture. The daily returns Friday with financial regulation.
CyberEyeQ — Actionable Regulatory Intelligence · [email protected]