Public sector use of AI risks undermining transparency - but better procurement could re-establish trust, according to a new government report.
The Artificial Intelligence and Public Standards review by the Committee on Standards in Public Life found the UK government should use its purchasing power to ensure private companies developing AI solutions for the public sector make their products more transparent.
It said a current lack of information about how the government uses AI risks obscuring the chain of organisational accountability, undermining attribution of responsibility for key decisions and could inhibit public officials explaining how decisions were reached using AI data.
“Public sector organisations are not sufficiently transparent about their use of AI and it is too difficult to find out where machine learning is currently being used in government,” it said.
The report recommended public bodies set ethical standards early in the procurement process, including making sure they are explicitly written into tenders and contractual arrangements.
The Office for AI, the Government Digital Service (GDS), and the Alan Turing Institute have already jointly issued draft guidelines on AI procurement.
However the Centre for Data Ethics and Innovation is also advising on regulation and AI use, making it difficult for companies know which set of guidance had priority.
The report said: “Multiple sets of ethical principles are confusing and the application of each is unclear.” It added: “Public sector guidance is not yet widely used and public officials with no AI expertise may find it difficult to understand and comply with.”
The report found procurement could play a key role in making public sector AI use more transparent, by ensuring products and services were explainable when they were contracted.
Software companies developing AI technology told the committee they had the capability to make their products and services more explainable, but they were rarely asked to do so by those procuring technology for the public sector.
Requirements for technical transparency are not usually included in procurement tenders and contracts. “To achieve this, explainability needs to be considered in the early stages of project development and design, and during procurement processes, by those commissioning the technology for use in the public sector,” said the report.
It also argued procurement processes should consider whether the AI products meet ethical standards and should also consider the ethical behaviour and culture of the supplier.
“Artificial intelligence has the potential to revolutionise the delivery of public services, creating an opportunity for more innovative and efficient public service delivery,” the report added.