As the EU AI Act enters its implementation phase, regulatory attention is moving 'from the model to the metal'. In a draft report circulated for consultation in July 2026, the European Commission's AI Office introduced the concept of 'Compute Governance'.
The proposal explores mechanisms to track and audit the use of large-scale AI training clusters defined as exceeding a certain computational threshold. The stated aim is to ensure that immense compute power is not used to train potentially harmful 'dual-use' foundation models without appropriate oversight.
The idea has drawn a mixed response across the Channel. The UK's Department for Science, Innovation and Technology issued a statement reiterating its 'pro-innovation' stance and cautioning against measures that could stifle research or create a competitive disadvantage, favouring voluntary safety commitments from data-centre operators and cloud providers over prescriptive legislation.
The divergence highlights a widening split in AI hardware regulation between the EU's rules-based approach and the UK's lighter-touch philosophy — a compliance landscape that multinational operators will have to navigate carefully.
Source: European Commission AI Office (draft consultation), July 2026.
