PPN 02/24: the original AI transparency note for procurement, explained

The key facts

• PPN 02/24 is an information note issued in March 2024, applying to all central government departments, their executive agencies and non-departmental public bodies. Other public sector contracting authorities may wish to apply it.
• It was the first procurement policy note dedicated to transparency of AI use, issued under the previous procurement regime. Its successor in the Procurement Act 2023 suite, PPN 017, directs buyers back to PPN 02/24 for procurements commenced before 24 February 2025.
• Suppliers' use of AI to develop bids is not prohibited during the commercial process. The note compares it to a supplier using a professional bid writer.
• The note provides three example AI disclosure questions for the invitation to tender. They should not be scored or taken into account in tender evaluation, and are for information only.
• Buyers should put proportionate controls in place so bidders do not use confidential contracting authority information as training data for AI systems or large language models.
• Where AI has been used in a tender, additional proportionate due diligence may be required, such as site visits, clarification questions or supplier presentations.
• Buyers should plan for a general increase in activity, including more tender responses and clarification questions, and consider allowing more time in the procurement.
• Where use of AI by suppliers raises national security concerns, commercial teams should engage information assurance and security colleagues before launching the procurement.

What PPN 02/24 is and where it now sits

PPN 02/24, issued in March 2024, was the government's first procurement policy note dedicated to the transparency of AI use in procurement. Its concern is simple: AI systems, tools and products are a rapidly growing and evolving market, government is adopting them, and the risks and opportunities need managing as part of commercial activity. It is an information note: guidance to apply, rather than a mandated set of actions.

It applies to all central government departments, their executive agencies and non-departmental public bodies, described in the note as in-scope organisations, and other public sector contracting authorities may wish to apply the same approach. It is aimed squarely at those in commercial, procurement and contract management roles.

For context on numbering: this note was issued under the previous procurement regime, and its content was later republished in the Procurement Act 2023 suite as PPN 017. PPN 017 itself tells buyers to refer to PPN 02/24 for procurements commenced and contracts awarded before 24 February 2025. So PPN 02/24 remains the operative reference for older procurements and live contracts let under the earlier rules, while the substance of the guidance is the same in both notes.

The buyer playbook the note established

The starting point is permissive and pragmatic. There are potential benefits to suppliers using AI to develop bids, including the ability to bid for more public contracts, and suppliers' use of AI is not prohibited during the commercial process. The note draws a deliberate comparison: the risks should be understood in the same way as when a bidder has used a professional bid writer.

Around that starting point, the note sets out a series of steps. Ask suppliers to disclose their use of AI in creating their tender. Put proportionate controls in place so bidders do not use confidential contracting authority information, or information not already in the public domain, as training data for AI systems, for example using confidential government tender documents to train large language models to write future tender responses. Undertake proportionate due diligence where AI has been used, which could include site visits, clarification questions or supplier presentations, to establish the accuracy, robustness and credibility of tenders and to confirm the supplier has the capacity and capability to deliver.

The note also asks buyers to plan ahead operationally: expect a general increase in activity as AI improves suppliers' bid writing capacity, with more clarification questions and more tender responses; consider allowing more time in the procurement; and align more closely with internal customers and delivery teams who understand what AI means for the subject matter of the contract. Where the use of AI by suppliers raises national security concerns, commercial teams should engage information assurance and security colleagues before launching the procurement.

The disclosure questions in detail

Annex B of the note provides three example disclosure questions for the invitation to tender. The first asks whether AI or machine learning tools, including large language models, were used to assist any part of the tender submission, with a request for details and a confirmation that AI generated content has been checked and verified for accuracy. The second asks the supplier to detail any instances where such tools were used to generate written content or support the bid. The third looks at delivery rather than bidding: are AI or machine learning technologies used as part of the products or services the supplier intends to provide, and if so, how are they integrated.

The rules around these questions matter as much as the questions. They should not be scored or taken into account in tender evaluation: they are for information only, feeding commercial strategy, due diligence planning and risk management. Contracting authorities can still ask and evaluate further AI questions specific to their requirements, provided they comply with procurement law and the approach to scoring is set out in the procurement documents. And authorities must not discriminate against particular suppliers either in using the questions or in interpreting the answers.

The delivery question is the sleeper. The note observes that AI and machine learning are increasingly prevalent in the delivery of services not sold as AI services, and illustrates the point with a video conferencing procurement where the supplier offers generative AI transcription or live translation. The recommended response is contractual: be clear about how data from the service will be managed, including that it is not used for training AI models unless specifically agreed and approved in writing.

Why the note matters: accuracy without understanding

The background section explains the underlying risk in plain terms. Large language models are trained to predict a statistically plausible string of text. Statistical plausibility is not factual accuracy: content created with LLM support can include statements, facts or references that appear credible but are false, and because the model has no contextual understanding of the question or its own answer, it cannot identify or correct its errors. The note also stresses, citing the government's Data Ethics Framework, that decisions should be made with the support of AI systems, not a reliance upon them.

For evaluators, the consequence is a change of habit rather than a change of law: treat impressive prose as a claim to be verified, use clarifications and supporting documentation exactly as you would for any ambiguity in a tender, and give the procurement timetable room for that verification. The note's annex of guidance, from the Guidelines for AI Procurement to the Generative AI Framework for HMG, points teams to the wider reading, but the operational message fits in a sentence: do not ban the tools, and do not trust them either.

How eSourcing Data helps

The disclosure approach in PPN 02/24 depends on asking consistent questions and keeping the answers out of the scores. eSourcing Data lets teams build the disclosure questions into tender documentation as unscored information questions, separate from evaluated criteria, so the distinction the note requires is enforced by the structure of the exercise rather than by memory.

The due diligence the note recommends, clarifications, presentations, supporting evidence, runs through the platform's messaging and evaluation workflows, which means every exchange with every supplier sits in the audit trail. If a decision is later questioned, the record shows AI disclosures were handled consistently and without discrimination.

For teams managing procurements under both the old and new regimes, a single structured record of which rules each procurement was commenced under, and which guidance applied, is exactly the kind of housekeeping that prevents confusion two years later.

What to do about it

1. 1Confirm which regime each live procurement was commenced under, and apply PPN 02/24 to those commenced before 24 February 2025.
2. 2Add the example disclosure questions to your invitation to tender as unscored, information-only questions.
3. 3Include proportionate controls preventing confidential authority information from being used as AI training data.
4. 4Verify AI-assisted tenders through proportionate due diligence: clarifications, supporting documentation, site visits or presentations.
5. 5Ask the delivery-side question: will AI feature in the service itself, and on what contractual terms is service data managed.
6. 6Allow more time in procurement timetables and resource evaluation for higher volumes of responses and clarifications.
7. 7Engage information assurance and security colleagues before launch where AI use raises national security concerns.