The UK Cabinet Office Says Suppliers Can Be Asked to Disclose AI Use in Tenders

Can a supplier use AI to develop its bid for a UK central government contract? A Cabinet Office note answers. Suppliers' use of AI "is not prohibited during the commercial process," it says. But steps should be taken to understand the risks.

The note is Procurement Policy Note 017, on improving transparency of AI use in procurement. It was originally issued in November 2023 and updated in February 2025. It applies to all central government departments, their executive agencies and non-departmental public bodies. Other public sector contracting authorities "may wish to apply" its approach, it says.

The update does not constitute a change in policy or a new call for action, the note says. But in-scope organizations should continue to apply any ongoing obligations in it. They should note its provisions from February 24, 2025. Parts of the note follow, not the whole.

AI in the Bid

The note says there are potential benefits to suppliers using AI to develop their bids. Doing so enables them to bid for a greater number of public contracts, it says.

Steps should be taken to understand the risks of AI tools in this context, the note says, as would be the case if the supplier had used a bid writer.

The note's background section says content created with the support of large language models may include inaccurate or misleading statements. Statements, facts or references may appear plausible but be in fact false. The models are trained to predict a "statistically plausible" string of text, the note says. That does not necessarily mean the statements are factually accurate, it adds.

Care must be taken, the note says, in the use of such models.

Steps the Note Lists

The note lists steps that understanding the risks may include. Some of them follow. One is asking suppliers to disclose their use of AI in the creation of their tender.

Another concerns confidential information. The note speaks of proportionate controls. They are to ensure suppliers do not use confidential contracting authority information, or information not already in the public domain, as training data for AI systems. Its example is confidential government tender documents used to train AI or large language models to create future tender responses.

The note also lists appropriate and proportionate due diligence. If suppliers use AI tools to create tender responses, it says, additional due diligence may be required to ensure suppliers have the appropriate capacity and capability to fulfill the requirements of the contract. It should be proportionate to any additional specific risk posed by the use of AI. It could include site visits, clarification questions or supplier presentations.

What does a well-written tender prove, once a tool can help write it?

The Disclosure Questions

Annex B of the note gives example disclosure questions. A disclosure question can be added to the invitation to tender, it says. It requires suppliers to disclose their use of AI when responding to the tender questions, or as part of their proposed delivery of the service.

The example questions "may be helpful," the annex says. One asks whether the supplier has used AI or machine learning tools, including large language models, to assist in any part of its tender submission. It asks for details. The same example also asks for a confirmation. Where AI tools have been used to support the generation of tender responses, the supplier is asked to confirm that they "have been checked and verified for accuracy." The answer is yes or no.

The annex's questions should not be scored or taken into account when assessing a tender, it says. They should be used for information only. Contracting authorities can still ask and evaluate further relevant questions about the use of AI as part of their award process, where those are specific to their requirements and comply with procurement law. And the note says it is important that contracting authorities do not discriminate against particular suppliers in the use of the questions, or in the interpretation of supplier responses.

The confirmation is a yes or a no. Who did the checking, and against what, the example questions do not ask.

AI in the Service

AI and machine learning are becoming increasingly prevalent in the delivery of what the note calls "non-AI" services, it says. Where AI is likely to be used in the delivery of a service, commercial teams "may wish to require" suppliers to declare this and provide further details. That will enable the teams to consider any additional due diligence or contractual amendments, the note says.

It gives an illustrative example: a contracting authority seeking a new video conferencing suite. A supplier may be able to provide a service that includes meeting transcription, or live translation using generative AI. Additional consideration should then be given to how the meeting records will be used, the note says, for example for training further AI models. It should also be given to whether the information gathered is subject to data classification policies.

Departments will likely want contract terms that make clear the data is appropriately managed, the note says. Its example of such a term: the data is not used for training purposes unless this has been "specifically agreed and approved in writing by the client."