The public sector needs to set rules around AI procurement as existing frameworks are “no longer fit for purpose”, according to a report.
A report, which includes policy recommendations to improve AI procurement in the UK, has warned that “existing frameworks for procurement and implementation are no longer fit for purpose and ill-equipped to provide guidance on the democratic use of AI”.
The report said: “To harness artificial intelligence in the public service, governments need to take decisions about their procurement and development.
“Lacking guidelines and best practices, governments are at risk of being sold unsuitable technologies by consulting and IT firms or developing inadequate products in-house.”
It recommended that public agencies needed to be trained in AI procurement due to its complexities and when buying from external suppliers questions need to be considered around “the design, usability, and aptitude of AI alongside other considerations around budget and timelines”.
“AI procurement should not proceed without a full understanding of data provenance, modelled outputs, and the governance structure of the firms producing the AI,” said the report.
The report was published by The Oxford Commission on AI and Good Governance, led by the Oxford Internet Institute. The commission has been formed to assess key challenges surrounding the adoption of AI technology in the public sector, including design, procurement, implementation and accountability, and provide recommendations to the government over the next 18 months.
Risks, benefits and whether the technologies operate “legally, ethically, and inclusively” also need to be assessed. “Complex technologies and their impacts are extremely difficult to evaluate and to conduct due diligence for,” said the report.
Public procurement departments need to ensure they are engaging effectively with a wide market of AI vendors, including big tech, startups, and research institutions, as “many government departments and agencies – especially at the local level – remain uncertain about how to engage with artificial intelligence and other automated decision support technologies”.
☛ Want to stay up to date with the news? Sign up to our daily bulletin.