What’s the issue with HMRC’s use of AI to assess R&D tax relief claims?

HM Revenue and Customs (HMRC) has recently come under fire amid allegations that artificial intelligence (AI) has been used, without disclosure, to assess the validity of R&D tax relief claims.

This revelation has raised serious questions about the reliability, security and fairness of the claims review process.

We have supported many clients in adapting to the changes made to R&D tax relief.

While it is widely accepted that stronger checks were needed as compliance rates are a consistent issue, the idea that AI, particularly generic systems like ChatGPT, may have played a role behind the scenes has understandably caused concern.

Was AI used in R&D tax relief claims reviews?

According to HMRC, no AI is currently being used to assess R&D tax relief claims.

Officially, every claim is reviewed by a human agent.

However, whistleblowers say otherwise.

They allege that some HMRC agents did use AI tools, specifically, Large Language Models (LLMs), to help assess claims.

These same sources also suggest this practice has now been stopped, but crucially, HMRC only addressed the issue after being forced to by a Freedom of Information (FOI) request.

Why is that a problem?

Given the confusion around the topic, it is no surprise that business owners like yourself may be wondering what the problem is with HMRC’s use of AI.

The main problems involve security risks, inaccuracy and unfair rejections and a lack of transparency.

Generic AI tools like ChatGPT are not built for handling sensitive taxpayer data.

If HMRC staff input claim details into an unsecured AI tool, it could violate data protection laws.

Likewise, LLMs are not trained to understand the complexities of R&D tax law or niche technical projects.

If AI has influenced decision-making, there is a real risk that valid claims were incorrectly flagged or rejected.

Additionally, if HMRC is using AI tools without disclosing it, claimants are being left in the dark about how their submissions are being assessed and whether the process is fair.

What are the risks to your R&D claim?

If your claim was reviewed during the period when AI tools may have been used, there is a chance that your claim was assessed based on inaccurate or misunderstood information.

This could have contributed to an overall issue of delays and further scrutiny due to inconsistent review processes.

More concerningly is the risk that sensitive data may have been incorrectly inputted into an LLM without your knowledge.

What can you do about it?

This situation highlights why it is more important than ever to have experienced professionals supporting your R&D claim.

We ensure every claim is technically robust and aligned with the latest HMRC guidance.

To do this, we work proactively with clients to identify all qualifying expenditure, no guesswork, no missed opportunities.

Where enquiries or challenges may arise, we support you in getting a resolution, especially if your claim has been unfairly flagged.

What could all this mean for the future of R&D tax relief claims?

While HMRC has not confirmed the use of AI in processing R&D claims, the patterns emerging raise serious questions about how claims are being reviewed behind the scenes.

HMRC must now be fully transparent about what has happened and how they will prevent similar issues from recurring.

Businesses deserve clarity and fairness when navigating a system that can significantly impact their ability to innovate.

In the meantime, companies must take proactive steps to protect themselves.

With increased scrutiny and a more unpredictable review process, submitting a claim that is thorough, well-evidenced and fully compliant is more important than ever.

If you are unsure whether your R&D claim could be at risk, or you simply want peace of mind that your next submission stands up to scrutiny, then get in touch with our team.

We’re here to help you navigate the system confidently and correctly.

For R&D tax relief support you can trust, speak to our team today!

Stephanie Hurst