Artificial intelligence (AI) specific considerations for partnering with a supplier
For forces going down the route of partnering with a supplier, the Data-driven technologies authorised professional practice (APP) emphasises the need for an informed and strategic approach to procurement, and signposts resources that will help you. The following considerations are specific to artificial intelligence (AI).
Legality, safety and ethics
The PROBabLE Futures and National Police Chiefs' Council (NPCC) Responsible AI checklist is the recommended checklist for forces to use when assessing the responsible nature of AI technologies or the specific application of that technology. It brings together technical, operational, legal and ethical aspects, and it is living document that will be updated periodically.
The checklist is grounded in the principles set out in the NPCC AI Covenant and the Data ethics APP. It includes case studies that show how to apply it, with examples of product features and results that constitute red, orange and green flags.
Alongside the checklist, you should also work through the issues flagged in your information security team (IST)-led risk assessment, grounded in PDS standards.
Being a discerning customer
When dealing with prospective suppliers, you should engage your force’s procurement expertise to establish the best routes to market and strategies for selecting and working suppliers. It is also helpful to be equipped with:
- a clear problem statement
- a clear understanding of budget and an awareness of the potential for hidden costs
- an assessment from your IT department on what data is available and the state it is in
- knowledge of which systems an AI tool or system will interoperate with
- a clear understanding of the balance of risk you want to agree with the supplier. Which risks does it makes sense for your force to own, and which should be owned by the supplier?
- strategies for avoiding vendor lock-in. These could include:
- basing the relationship on rolling short-term contracts
- requiring that tools are supplier-agnostic and technology-agnostic
- requiring that intellectual property implications are fully assessed (for example, if any new algorithms are being generated or if an existing one is being retrained on your force’s data, which could give rise to new intellectual property)
- request open source and open standards (standards that are freely available for adoption, implementation and updates, such as XML, SQL and HTML)
- stipulate in the contract when their involvement will end, as well as a requirement for them to upskill your force and to transfer any required knowledge before that point
- ensuring that any ongoing relationship is limited to maintenance and repairing faults
For further guidance go to GOV.UK Guidelines for AI procurement.