A data clean room fits into an agent-negotiated deal as a separate, rule-bound place where buyer and seller data meet, not as something either agent controls. Agents can request what its rules permit: an overlap count before the deal, a matched segment during it, an aggregated outcome report after it. They cannot decide whether the data may be used in the first place. Under UK GDPR, that decision, including the lawful basis and any consent it relies on, stays with the data controller.

TL;DR. A data clean room lets two or more organisations match and analyse their data under agreed rules without either seeing the other's records row by row. In an agent-negotiated deal, clean rooms are useful at three points: sizing an audience before a deal, activating a matched segment during it, and measuring outcomes afterwards. An agent can only submit queries the clean room permits and use outputs its mandate allows. The lawful basis for processing personal data, and any consent, remains the responsibility of each data controller, and the deal record should show that a person confirmed it.

What is a data clean room?

A data clean room is a controlled environment in which two or more parties bring data, match it on agreed keys, run approved analyses and receive outputs, usually in aggregate, without either party seeing the other's underlying records. Typical controls include an approved list of queries, minimum group sizes below which results are suppressed, limits on how outputs can be combined, and review of what leaves the environment.

Clean rooms are not one standard design, and they differ in who operates them and which controls apply. For a deal, the useful question is not "is there a clean room?" but "what does this clean room permit, and who agreed those rules?"

Where do clean rooms fit in an agent-negotiated deal?

Clean rooms can support an agent-negotiated deal at three stages.

Before the deal. A buyer agent asks how much of the advertiser's customer base the publisher reaches. The clean room returns an aggregate overlap figure, which both agents can use as an input to price and volume.
During the deal. A matched segment is created inside the clean room and made available for targeting. The deal terms refer to the segment by identifier, so no personal data passes through the negotiation itself.
After the deal. The advertiser's conversion data and the publisher's exposure data are matched to produce an aggregated outcome report, which is especially relevant when the price depends on outcomes.

What can an agent do with a clean room, and what can it not?

An agent can submit queries the clean room permits, read aggregated outputs, use them as negotiation inputs and reference segment identifiers in proposed terms.

An agent cannot widen the purposes the data was brought in for, establish a lawful basis, or decide that a new use is acceptable. It also should not be allowed to run sequences of narrow queries whose outputs, combined, could single out individuals. A clean room's thresholds help, but the buyer's mandate should forbid the behaviour as well, and route any query outside the approved set to a person.

A query an agent runs is processing carried out for the organisation that deployed it, which remains accountable for it.

Who is responsible for consent and lawful basis under UK GDPR?

Under UK GDPR, the data controller, the organisation that decides why and how personal data is processed, must establish a lawful basis before processing begins. In a typical clean room arrangement, each party bringing personal data is a controller for its own data. Where parties jointly decide the purposes and means of a particular analysis, they may be joint controllers. An organisation operating the clean room on the parties' instructions will usually act as a processor, although the contracts decide the actual roles.

Several points follow for agent-negotiated deals.

Hashed or pseudonymised identifiers used for matching are still personal data under UK GDPR, because they can be linked back to individuals with additional information.
Where consent is the lawful basis, it has to cover the use being made of the data. Where cookies or similar technologies are involved, the Privacy and Electronic Communications Regulations (PECR) apply alongside UK GDPR.
Neither a clean room nor an agent transfers these obligations. A buyer agent negotiating against a segment is relying on its principal, and the publisher, having done this work.

The ICO publishes guidance on these roles, and organisations should take legal advice on their own arrangements.

What should the deal record say about clean room use?

The agreed terms of a deal that relies on a clean room should cover:

Which clean room environment is used and whose data enters it.
The permitted purposes and queries for this deal.
The segment identifiers to be activated, rather than any personal data.
Which clean room output is the source of truth for measurement or billing.
Minimum aggregation thresholds and limits on further analysis.
Retention and deletion of matched data after the campaign.
A confirmation from each controller, given by a person, that a lawful basis is in place for its data.

What does this look like in a hypothetical deal?

For example, a hypothetical retailer wants to reach its lapsed customers on a news publisher's site. The buyer agent requests an overlap query, and the clean room returns an aggregate figure of 240,000 matched users. The publisher's agent proposes a private marketplace deal against that segment at a stated CPM for a four-week flight, and the buyer agent accepts within its mandate, with both companies' data protection leads having confirmed the lawful basis beforehand.

Midway through, the buyer agent asks for results by postcode district. Some groups fall below the clean room's minimum threshold, the request is refused, and the agent escalates to a person rather than trying narrower combinations. The final report, matched in aggregate inside the clean room, is the named source for billing, and both parties' records refer to the same report.

How do AAMP and AdCP handle audiences?

AAMP, IAB Tech Lab's Agentic Advertising Management Protocols, includes agentic-audiences, an open protocol for interoperability across advertising agents working with audiences. AdCP, the separate protocol published by AgenticAdvertising.org, includes sync_audiences for bringing in first-party audiences and get_signals and activate_signal for targeting data. Protocols like these standardise how audience signals move between agents. They do not establish the lawful basis for the data behind those signals.

The clean room governs the data. The terms the parties agree on the strength of it need a record both sides hold. Alkimi, a neutral agentic advertising marketplace working within IAB Tech Lab's AAMP framework, verifies that each connecting agent is authorised by the buyer or publisher it represents, and writes agreed terms to a DealSheet that both parties hold and either can retrieve independently, with people keeping approval rights over each deal.

Frequently asked questions

What is a data clean room in advertising?

A controlled environment where advertisers, publishers or other parties match and analyse their data under agreed rules, receiving outputs, usually aggregated, without seeing each other's underlying records.

Can an AI agent use a data clean room?

Yes, within the clean room's permitted queries and the agent's mandate. The agent can request approved analyses and use aggregated outputs, but it cannot extend the purposes the data may be used for.

Does a clean room make data anonymous?

Not automatically. Identifiers used for matching, including hashed ones, are still personal data under UK GDPR. Whether a given output is anonymous depends on how it is aggregated and what else it can be combined with.

Who is responsible for consent when agents use clean room data?

The data controller. Each party bringing personal data must establish its own lawful basis, and any consent, before processing. Using an agent does not change that.

How Do Data Clean Rooms Fit Into Agent-Negotiated Deals?