Public‑sector buyers must look beyond brochure names, as new CMA guidance on agentic ai in public sector raises compliance stakes.
CMA’s March research paper
On 9 March 2026 the UK Competition and Markets Authority released a research paper examining how agentic artificial intelligence may affect consumers. The paper, summarised by ReedSmith, flags a suite of risks – dark‑pattern manipulation, erosion of consumer agency and the possibility of “agentic collusion” where autonomous systems coordinate to influence markets.
Compliance obligations and potential penalties
The CMA’s accompanying compliance guidance requires traders, including public‑sector bodies that deploy AI agents, to disclose the use of such agents to customers and to avoid overstating system capabilities. Breaches of UK consumer‑protection law can attract fines of up to 10 % of global annual turnover per breach, as ReedSmith notes.
Timeline of regulatory activity
Two CMA publications bookend the recent focus on agentic AI. On 4 March 2026 the regulator issued a short piece titled “AI and Collusion: Frontiers, Opportunities and Challenges”. A week later, on 9 March, the full research paper and guidance were released. Both documents are referenced in the ReedSmith analysis.
What the guidance means for public‑sector procurement
Public‑sector procurement teams now have to embed the CMA’s disclosure rules into tender specifications. While the guidance does not prescribe a specific method‑versus‑results comparison, it makes clear that any marketing claim about an AI system’s autonomy must be substantiated, and that misleading statements could trigger the 10 % turnover fine.
Broader regulatory context
The UK Information Commissioner’s Office (ICO) has also published a Tech Futures report on agentic AI. It outlines data‑protection risks and sketches four possible adoption scenarios for organisations over the next two to five years. Although the ICO report does not address procurement directly, it reinforces the need for robust impact assessments before deploying autonomous agents in public services.
Industry perspective
Elsewhen’s analysis of agentic AI stresses the technology’s potential to reshape public‑sector workflows – from education planning to hospital discharge – by automating tasks that previously required hours of manual effort. The report, however, stops short of offering procurement guidance, underscoring the regulatory gap that the CMA’s new guidance begins to fill.
Further reading
- ReedSmith analysis of CMA activity
- ICO Tech Futures report on agentic AI
- How regulators are shaping AI policy in 2026
- Public‑sector technology procurement trends
Key dates and documents
| Date | Document | Source |
|---|---|---|
| 4 Mar 2026 | “AI and Collusion: Frontiers, Opportunities and Challenges” (short piece) | ReedSmith |
| 9 Mar 2026 | Research paper and compliance guidance on agentic AI | ReedSmith |
| Source: ReedSmith excerpts cited in the research packet. | ||
The CMA’s guidance marks the first explicit regulatory signal that public‑sector buyers must treat autonomous AI agents as subject to existing consumer‑protection rules. What remains unclear is whether future guidance will prescribe a formal method‑vs‑results assessment framework for procurement teams. Until then, agencies are advised to embed disclosure clauses, conduct risk‑based impact assessments and monitor the evolving regulatory landscape.