Agentic AI is developing rapidly and raises competition issues that cut across the Committee's previous work on algorithms, AI and digital markets. There are important open questions about how existing competition frameworks apply to autonomous AI agents, the respective roles of ex-ante and ex post tools, and the role of international co-operation, as well as the role of regulations and other policy areas. At the same time, agentic AI technologies and business models are not yet mature. How they develop will greatly affect the functioning of competition and the role, if any, competition authorities should play.1
Nonetheless, preserving and promoting effective competition in agentic AI services is important. Competition authorities should therefore consider developments closely and build relevant technical capacity. Existing competition enforcement and merger control tools, market studies, advocacy and, where applicable, ex ante regulatory tools may also have a role in supporting competitive outcomes, alongside domestic and international co-operation.
In monitoring developments, competition authorities should be prepared to use existing competition enforcement tools if there are potential breaches of competition law, both in terms of the supply of agentic AI and its effects on competition downstream. Assessing the evolution of these markets and any emerging market power can help authorities identify competition concerns and respond with enforcement if necessary. In particular, competition authorities may closely consider vertical and other relationships, and the potential exercise of market power from adjacent markets into agentic AI, such as leveraging positions in existing distribution models with pre-installation or default settings or bundling services through integration (Bostoen and Krämer, 2026[16]). Existing rules on co-ordination and information sharing may also become relevant where firms delegate commercially sensitive decisions to agentic AI systems, rely on common providers or permit systems acting on their behalf to interact. As such developments are still in their infancy, there may be value in tracking their progress.
Effective enforcement may also depend on the auditability of agentic AI systems. Records of relevant instructions, information flows, communications and actions may help authorities reconstruct conduct, even where the internal operation of the underlying model cannot be fully explained. Competition authorities may wish to consider how the tools at their disposal, including, as discussed below, co-operation across policy fields, could help promote explainability and auditability of agentic AI systems. Given the speed of developments in agentic AI, competition authorities may also wish to consider the use of interim measures where appropriate, as illustrated by the European Commission case discussed in Box 4.
Merger control will also likely be a key tool for competition authorities in promoting and preserving competition in agentic AI markets. As noted in Box 3, there has been significant discussion on the role of partnerships, acqui-hires and investments in the AI supply chain and whether these could lessen competition. The emergence of agentic AI adds further activities and relationships that may become competitively significant, including orchestration, access to tools and services, integration with existing software and distribution to users. Transactions involving these activities may therefore warrant consideration alongside those involving established AI inputs such as models, data and compute.
More broadly, given the fast-moving nature of agentic AI, and potential difficulty in identifying breaches of competition law from the outset, competition authorities may wish to proactively use market studies and advocacy to increase their understanding of these markets. This could include examining emerging business models and patterns of adoption and delegation, the extent to which agentic AI systems become important gateways between users and suppliers, barriers to switching and interoperability, and whether commercial relationships affect how agents select between competing offers. Market studies could also help identify whether any emerging concerns are best addressed through competition enforcement, advocacy or other policy tools. To assist authorities in this task, they may consider proactive engagement with agentic AI suppliers. Advocacy and engagement with suppliers could also explore whether competition risks can be mitigated through the training and design of models and agentic AI systems, for example by reducing the likelihood that systems facilitate anticompetitive conduct or inappropriate exchanges of competitively sensitive information. The effectiveness of such measures remains uncertain, but early engagement could help raise awareness of competition risks as systems are developed and deployed.
Over the past few years, certain jurisdictions have adopted ex-ante digital competition regulations, such as the Digital Markets Act in the European Union (OECD, 2024[68]) and the Digital Markets, Competition and Consumers Act in the United Kingdom. As these regimes were generally created with previous digital services in mind, questions may arise as to how they apply or adapt as new technologies and services emerge. This may be particularly relevant as agentic AI capabilities are integrated into existing digital ecosystems, as well as if agentic AI systems themselves develop into important gateways between users and other businesses or services (Bostoen and Krämer, 2026[16]).
The wider regulatory and policy environment may also affect how competition in agentic AI develops. This could call on advocacy if there are potential risks to competition from regulations that appear unmerited. Some authorities may also wish to consider the interaction with industrial policy and the potential for pro-competitive interventions, as well as being aware of the broader policy climate. This includes being aware of the broader risks and policy discussion on AI, as well as closer considerations such as national security and foreign investment screening. This could be particularly relevant for mergers, with more general implications for defining product and geographic markets and for barriers to entry.2
Co-operation is likely to be a key tool for competition authorities. As a new technology that will affect a range of policy areas, co-operation across governments and regulators is likely to be valuable for competition authorities to understand how they are approaching developments and any shared challenges.3 For example, consumer protection is an area that some have argued may need to adapt to the realities of agentic AI (Busch, 2026[69]), but many other areas are likely to be relevant, such as privacy and data protection, communications regulation, amongst others. Co-operation with other policy areas could also be useful to identify and tackle emerging issues that may not fall solely or squarely within the remit of competition law, but nonetheless have the potential to affect competition. This could include advocacy to ensure that regulatory systems are appropriately flexible to encourage the emergence of pro-competitive benefits from agentic AI.4 Establishing dialogue at a national level could be useful, including pursuing areas of general interest, such as promoting the explainability of agentic AI systems.5
As well as domestic co-operation, international co-operation is likely to remain a valuable tool for competition authorities. Agentic AI is unlikely to be confined by national borders and thus, were competition issues to emerge, these would likely be shared by more than one jurisdiction. Further, sharing knowledge and expertise can reduce duplication and aid authorities in understanding these emerging markets and monitoring them for any risks to competition. Bodies such as the OECD and International Competition Network are likely to be useful vehicles to facilitate and promote co-operation going forward.
Finally, while not covered in this paper, another dimension that is already on many competition authorities’ agendas, is the ability for them to use agentic AI tools to improve the effectiveness and efficiency of their work. While this paper has focussed on the potential risks to competition from and in agentic AI, it has also highlighted its potential benefits for productivity and innovation. Competition authorities can also benefit from this, potentially increasing their ability to monitor and detect competition issues, as well as improving their overall efficiency and effectiveness (May, 2026[70]).6