[{"data":1,"prerenderedAt":29},["ShallowReactive",2],{"nr-en-eu-commission-can-audit-gpai-models-and-demand-documents":3},{"slug":4,"title":5,"dek":6,"date":7,"time":8,"publishedAt":9,"updated":10,"updatedAt":10,"dateFmt":11,"updatedFmt":10,"kind":12,"tier":13,"author":14,"authorName":15,"topics":16,"tracker":21,"trackerLabel":22,"headlineStat":23,"image":24,"ogImage":25,"imageAlt":5,"csv":10,"minutes":26,"words":27,"html":28},"eu-commission-can-audit-gpai-models-and-demand-documents","EU AI Act: Commission can audit GPAI models and demand documents","Since 2 August 2026, the European Commission can audit and sanction providers of general-purpose AI (GPAI) models. New transparency rules and postponed deadlines supplement the framework.","2026-11-06","06:24","2026-11-06T06:24:00+01:00","","November 6, 2026","news","standard","ideal-syka","Ideal Syka",[17,18,19,20],"EU AI Act","GPAI","Regulation","Enforcement","\u002Feu-ai-act-fahrplan","EU AI Act timeline","Since 2 August 2026, sanctioning powers for GPAI providers","\u002Fnewsroom\u002Fimg\u002Feu-commission-can-audit-gpai-models-and-demand-documents.webp","\u002Fog-nr\u002Feu-commission-can-audit-gpai-models-and-demand-documents.en.png",2,462,"\u003Ch2>What has changed\u003C\u002Fh2>\n\u003Cp>According to BornCity, the European AI Office received far-reaching control and enforcement powers with the start of concrete enforcement of the EU AI Act on 2 August 2026. For general-purpose AI (GPAI) models, the European Commission can now demand documentation and direct model access for review. In case of established violations, it may order binding corrective measures and impose fines.\u003C\u002Fp>\n\u003Ch2>Deadlines for providers\u003C\u002Fh2>\n\u003Cp>The source names 2 August 2026 as the date from which the Commission has formal sanctioning powers over GPAI developers. According to the report, the basic obligations for these providers have applied since 2 August 2025. For providers who entered the market after that date, a transition period runs until 2 August 2027.\u003C\u002Fp>\n\u003Cp>According to BornCity, new transparency rules also took effect on that date: users must be clearly informed when they interact with certain AI systems. The same applies to systems for emotion recognition or biometric categorisation. Deepfakes and certain synthetic content must be labelled, and AI-generated data should carry a machine-readable marking. For generative systems already on the market before 2 August 2026, the source cites an implementation deadline in December.\u003C\u002Fp>\n\u003Cp>According to the report, the so-called Digital Omnibus readjusted the timelines for other segments in July 2026: obligations for high-risk AI were moved to 2 December 2027, and for AI components in physical products such as medical devices, toys and lifts to 2 August 2028.\u003C\u002Fp>\n\u003Ch2>Technical review and whistleblower portal\u003C\u002Fh2>\n\u003Cp>According to the source, at the end of August 2026 the Commission requested information from more than 30 unnamed AI companies, which is being evaluated by EU technology chief Henna Virkkunen. For the technical risk assessment, the AI Office and Directorate-General CNECT commissioned a consortium led by Capgemini, which includes Wavestone and the French testing institute LNE, according to the report. The focus, per the source, is on offensive cyber risks, such as the automated discovery of security vulnerabilities. Benchmark results or evaluation metrics of the contract were not published.\u003C\u002Fp>\n\u003Cp>According to BornCity, the AI Office has also launched a digital whistleblower portal through which concerns about dangerous or unlawful practices can be submitted anonymously in all EU official languages.\u003C\u002Fp>\n\u003Ch2>Voluntary code of conduct\u003C\u002Fh2>\n\u003Cp>According to the source, Anthropic, Google, IBM, Microsoft, Mistral AI and OpenAI signed a voluntary Commission code of conduct. According to the report, xAI participated only in the safety chapter.\u003C\u002Fp>\n\u003Ch2>Assessment\u003C\u002Fh2>\n\u003Cp>For companies that deploy or offer GPAI models, it could be relevant that documentation and access obligations are now enforceable. According to the source, it remains open how the Commission will concretely evaluate the requested information and which corrective measures it will order.\u003C\u002Fp>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fborncity.com\u002Fnews\u002Feu-ai-act-kommission-darf-gpai-modelle-pruefen-und-dokumente-verlangen\u002F\">BornCity\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cem>Editorially owned by \u003Ca href=\"\u002Fen\u002Fautor\u002Fideal-syka\">Ideal Syka\u003C\u002Fa>. Sources and method: \u003Ca href=\"\u002Fen\u002Fredaktion\">Newsroom & method\u003C\u002Fa>. Tips and corrections: \u003Ca href=\"mailto:ai@i6eal.de\">ai@i6eal.de\u003C\u002Fa>.\u003C\u002Fem>\u003C\u002Fp>\n",1793952073989]