Latest / Frontier models and safety
EU AI Office gains powers to enforce AI Act rules on general-purpose models
From 2 August 2026 the Commission's AI Office, with national authorities, began enforcing the AI Act, including obligations on general-purpose AI model providers that had applied since August 2025 and new transparency and deepfake-labelling duties. The Office can demand documentation, access models for evaluation, order mitigations and fine providers up to €15m or 3% of global turnover, though it said it would start with 'compliance dialogues'.
Why it matters
Frontier labs selling in Europe now face a regulator with evaluation access and fining power over their models, not just a voluntary code.
Line of Thought
Follow this story
Pick any item to keep going. Your path builds up above as a line you can share.
Curated lines through this story
Directly linked
Connections our researchers recorded
- DevelopmentEU publishes General-Purpose AI Code of Practice ahead of AI Act model duties10 Jul 2025 · Rule change · EUPuts into practice: Enforcement of the obligations the code was written for
What led here
Earlier developments on the same thread
- DevelopmentEU Digital Omnibus on AI enters force, delaying high-risk AI Act duties to 2027-2827 Jul 2026 · Rule change · EU
- DevelopmentEU pushes AI Act duties for hiring and workplace AI back to December 20278 Jul 2026 · Rule change · EU
- DevelopmentEU finalises code of practice on marking and labelling AI-generated content10 Jun 2026 · Policy · EU
- DevelopmentEU deal keeps AI medical devices under AI Act high-risk rules, delayed to 20287 May 2026 · Rule change · EU
What happened next
Later developments on the same thread
- DevelopmentOpenAI pauses frontier RL training over cyber risk after Hugging Face breach18 Aug 2026 · Statement · US
- DevelopmentGPT-6 Astra jumps to 62.7% on ARC-AGI-3, six months after models scored 0.5%3 Sep 2026 · Report · US
- DevelopmentGoogle DeepMind publishes SynthID Bio watermarking for AI-designed proteins in Nature30 Sep 2026 · Research · GB, US
- DevelopmentOpenAI starts watermarking ChatGPT and Codex text in the EU to meet AI Act rules5 Oct 2026 · Launch · EU
Same story elsewhere
What other countries and bodies did on this
- DevelopmentSouth Korea's AI Basic Act takes effect, with a one-year pause on fines22 Jan 2026 · Rule change · KR
- DevelopmentOpenAI launches GPT-6 Astra, first model it rates 'Critical' for cyber capability3 Sep 2026 · Model release · US
- DevelopmentNew York finalises RAISE Act, aligning frontier AI law closely with California27 Mar 2026 · Rule change · US-NY
- DevelopmentCalifornia enacts SB 53, first US state law on frontier AI transparency29 Sep 2025 · Rule change · US-CA
Rules in play
Laws and guidance this touches
- RuleEU AI ActEU · Partly in force · 13 Jun 2024
- RuleDigital Omnibus on AIEU · In force · 8 Jul 2026
- RuleGPAI Code of PracticeEU · In force · 10 Jul 2025
- RuleAI content marking and labelling codeEU · In force · 10 Jun 2026
- RuleAI Basic ActKR · In force · 21 Jan 2025
Threads by topic: Frontier models Safety testing Labelling AI content Deepfakes