AI safety and competition law: an analysis of the limits

Competition law can address economic concentration in AI, but cannot resolve every question of political influence, national security or system safety. That is the argument of the commentary “Where Competition Law Stops: Political Power, National Security and AI Safety”, published by Matrix Chambers. Its argument concerns the allocation of responsibilities: protecting rivalry between businesses differs from deciding whether to accept less competition to pursue safety or security objectives.

According to the commentary, the UK Competition and Markets Authority can order business divestitures following a market investigation under the Enterprise Act 2002, without establishing an infringement. The Digital Markets, Competition and Consumers Act 2024 offers another route through pro-competition interventions concerning firms designated with strategic market status. For the EU, the analysis points to draft merger guidelines published in April 2026, which would incorporate resilience and security of supply into the competitive assessment. The guidelines are described as proposed, rather than final.

On AI safety coordination, the commentary reports that bills introduced in the US Congress in July 2026 would create an antitrust exemption for coordination delaying or limiting the release of high-risk models, subject to advance notice and safeguards. The exemption is not described as already in force. The analysis concludes that legislatures and safety regulators should set the conditions for any coordination and be accountable for the trade-off with competition. This account of the legal framework remains based on the commentary: the legislative texts and draft guidelines have not been directly verified.