UK Gov Fails To Publish Use of AI in Public Sector – Why the Transparency Gap?

Key Takeaways

• In August, the U.K. government said it would require all public sector bodies to declare their use of AI tools.
• However, so far, the transparency initiative has been a failure.
• In three years, just nine entries have been made on the government’s algorithmic tools register.

When the U.K. government committed to publishing a register of AI tools used by public agencies in August, it was celebrated by transparency campaigners who were troubled by the steady creep of automated decision-making in the public sector.

Yet, although that creep shows no signs of slowing, there hasn’t been a single new entry on the algorithmic tools register  since the government said disclosure would be mandatory.

The Algorithmic Tools Register

Initially launched in 2021 as a voluntary initiative to encourage transparency, the algorithmic tools register acts as a repository where public bodies deploying AI can record details of the tools they use.

While some of the tools listed on the register are fairly run-of-the-mill AI applications like a recommendation engine used to aid navigation of the government website, others are more embedded in civic decision-making.

For example, the Food Standards Agency has reported that it uses an AI platform to predict which businesses might be at a higher risk of non-compliance with food hygiene regulations. This information is then used by authorities to help determine where to send inspectors.

No New Entries Despite Growing AI Adoption

Between 2021 and July 2024, nine public bodies registered AI tools on the government’s transparency platform.

However, since the government said it would make registration mandatory, there have been no new entries.

Meanwhile, the Public Law Project (PLP) has identified  55 automated decision-making systems used by the Home Office, Department for Work and Pensions (DWP), Ministry of Justice and Ministry of Defence, as well as several local authorities and police constabularies.

This figure has risen from 41 when the PLP register was launched in 2023. Given that the information it contains was gathered by independent researchers using freedom of information requests and public records, it also may not reveal a full and up-to-date picture.

Government Transparency Failures

In comments  reported by the Guardian, Peter Kyle, the secretary of state for science and technology, admitted the public sector “hasn’t taken seriously enough the need to be transparent in the way that the government uses algorithms.”

“I accept that if the government is using algorithms on behalf of the public, the public has a right to know,”

He added. “The only way to do that is to be transparent about their use.”

The most concerning aspect of the government’s transparency shortcomings is the failure of the Home Office and DWP to report any of the dozens of AI tools they use identified by PLP.

Names like the Home Office’s “Asylum Initial Decision Model” suggest the department has incorporated algorithmic decision-making into the processing of asylum claims, yet claimants haven’t been informed.

Meanwhile, a suite of fraud risk tools used by DWP recalls a 2023 welfare scandal in the Netherlands. In that case, Amnesty International  uncovered evidence that an AI tool used to process childcare benefit claims displayed unfair bias against racial minorities, who were disproportionately flagged as fraud risks.

Without more transparency, it will be difficult to hold public sector organizations accountable, eroding public trust and increasing the likelihood of illegal or unfair AI systems going unchecked.