Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
The UK tax authority has been accused of “a cover-up” of unauthorised use of artificial intelligence tools by its officers to reject tax credits for businesses carrying out research and development.
The controversy comes after HM Revenue & Customs in August lost a legal battle at a tax tribunal and was ordered to reveal whether it had used AI when making research and development credit judgments.
The tax authority responded last week in a letter, seen by the Financial Times, that said the R&D tax credits compliance team did not use generative AI as part of work on R&D tax relief claims.
“This technology was not approved for use in generating taxpayer letters,” it added.
However, two separate advisers have told the FT, based on conversations with HMRC staff, that, while it was never HMRC’s policy to use AI to respond to R&D claims, some individual caseworkers did so.
Advisers to companies seeking to obtain the R&D tax credit have raised concerns that some taxpayers might have faced penalties over R&D credit claims solely on the basis of AI assessments.
There have also been fears that, if officials entered details of tax credit claims into public large language models such as OpenAI’s ChatGPT, commercial confidentiality might have been compromised.
One person told the FT that “a number of people” had been disciplined for using AI in the preparation of responses to R&D claims in 2023. The individuals concerned were part of the department’s individuals and small business compliance directorate, which deals with the bulk of R&D claims.
Another person said caseworkers had used AI tools when corresponding with R&D claimants in the past but added: “They’re not using it now”.
The person added that all caseworkers had been trained on when and how to use the technology after HMRC’s leadership discovered the unauthorised use of generative AI on R&D tax correspondence by “the odd caseworker”.
The revelations about the caseworkers prompted Tom Elsbury, the tax expert who brought the successful transparency claim at the tax tribunal, to question the candour of the agency’s response to the tax tribunal ruling.
He said HMRC had worded its response “in a way which appears to try and divert attention from the issue by clarifying any ‘approved’ use of AI while leaving the point around unapproved use by caseworkers unaddressed.”
He added: “It seems like there is a bit of smoke and mirrors going on here”.
Richard Lewis of Pronovotech, an R&D claims company, who has previously raised questions about whether HMRC used AI to reject his clients’ claims, also said the agency was not being “totally transparent”.
He added: “It almost sounds like it’s a cover-up.”
In 2024, Lewis was told in a letter sent by HMRC in response to a complaint about the handling of an R&D claim: “We do not use artificial intelligence to prepare our correspondence”.
HMRC has put R&D tax credits under increased scrutiny in recent years after it uncovered huge levels of fraud and error.
However, professional bodies and tax experts have warned HMRC’s attempts to crack down on R&D tax credit abuse have gone too far. They have said that many businesses with genuine claims have been rejected and several have chosen not to apply for the credits.
In its response, HMRC conceded that, since the original freedom of information request on the issue in 2023, it had tested an internal R&D summariser tool in 2024 on “a small number of cases”. The cases were all ones where caseworker assessment and action had already been taken, it said.
When asked about the claims about individual caseworkers’ AI use, HMRC reiterated that it did not use the technology for R&D claims.
It said: “Where risks are identified, an enquiry is opened by a human who checks those claims, maintains customer contact throughout and makes the final decisions on that enquiry.”
It added: “Any staff found misusing AI would face disciplinary action.”
Read the full article here