Source: Cornerstone

Research and the Dynamic Competition Initiative convened the third annual Brussels Expert Forum on 9 June 2026, where three panels examined the European Commission’s new draft merger guidelines and their treatment of dynamic competition, innovation, data-based IP and AI. The discussion emphasized the Commission’s effort to delineate pro-competitive scale from anti-competitive market power while retaining the significant impediment to effective competition (SIEC) test as the binding legal standard. Panelists noted the draft’s structural convergence with the 2023 US merger guidelines, including a unified horizontal and non-horizontal approach, and an expanded focus on dynamic effects and non-price theories of harm, while also flagging divergences such as heightened attention to resilience and sustainability and references to a Commission “margin of appreciation.”

Debate centered on new or expanded analytical tools in the draft: the formal introduction of a “theory of benefit” alongside theories of harm, the innovation theory of harm, a standalone common-ownership theory positing a 5% threshold, and an “innovation shield” intended to protect acquisitions of small innovative companies from an SIEC finding. Panelists expressed skepticism about whether evidentiary standards for benefit theories can be symmetrical with harm theories in practice, concerns over operationalising innovation theories (including unresolved out-of-market efficiencies and unclear limiting principles), and institutional implications requiring greater Commission transparency to build durable case law. The forum also addressed implementation granularity — distinguishing elimination of potential competitors, loss of R&D-driven innovation competition, and loss of investment competition in network industries — urging parties to provide clear, provable transaction rationales early; and it explored IP’s elevated role in merger assessment, cautioning against mechanical reliance on aggregate IP metrics and highlighting transatlantic differences in AI training and copyright regimes.