Ad/Marketing Communication
This legal article/report forms part of my ongoing legal commentary on the use of artificial intelligence within the justice system. It supports my work in teaching, lecturing, and writing about AI and the law and is published to promote my practice. Not legal advice. Not Direct/Public Access. All instructions via clerks at Doughty Street Chambers. This legal article concerns AI Law.
The First-tier Tribunal’s decision in Get Onbord Ltd (in liquidation) v HMRC marks an important reminder of how the UK’s research and development (R&D) tax credit system can apply to software projects, particularly those incorporating artificial intelligence (AI). The judgment turned on whether the work undertaken by Get Onbord Ltd (“GOL”) amounted to “research and development” as defined in the legislation and guidelines. Ultimately, the Tribunal concluded that GOL’s project did qualify, emphasising the significance of AI-driven innovation in the sphere of R&D.
Background of the Case
Get Onbord Ltd was a start-up focused on creating an automated AI system for ‘know your client’ (KYC) verification and risk profiling. Its core goal was to develop software that outperformed manual processes for customer due diligence, especially within financial institutions that combat money laundering, fraud, and other financial crimes. GOL claimed R&D tax credits under section 1054 of the Corporation Tax Act 2009 (CTA 2009), but HMRC initially rejected the claim on grounds that there was no “advance in science or technology” beyond readily available or deducible methods.
Although GOL eventually went into liquidation, the Joint Liquidators continued the appeal, adopting the arguments already presented by the former director who testified before the Tribunal. The key question was whether GOL’s project met the statutory and regulatory definitions of “research and development.” These definitions are set out in CTA 2009, CTA 2010, the Income Tax Act 2007, and the BEIS Guidelines on the Meaning of Research and Development for Tax Purposes.
AI’s Key Role in the Project
- GOL’s system sought to replace extensive manual analysis with a holistic AI-driven approach. This included verifying identities across multiple data sources, integrating various big data streams, and reducing false positives through refined “unique-ing” algorithms.
- The project entailed writing large amounts of new code and testing different machine learning models. Although the team utilised existing open-source libraries, much of the work involved developing custom AI algorithms to adapt to fast-evolving fraud techniques, regulatory requirements, and risk management thresholds.
- GOL’s innovation lay in combining multiple data points, from sanctions lists to IP geolocation, to produce real-time decisions. The company used existing APIs and code libraries but built significant software layers on top to address uncertainties not trivially solved by off-the-shelf systems. This “holistic” approach was far more than a routine adaptation.
The Tribunal’s Reasoning and Conclusion
The relevant provisions in sections 1044–1054 of the Corporation Tax Act 2009 (CTA 2009), together with CTA 2010, require that expenditure on R&D must address genuine scientific or technological uncertainties and yield an advance in the field. According to the BEIS Guidelines, projects must go beyond merely deploying known technologies or engaging in routine adaptation, instead resolving uncertainties not readily solvable by existing knowledge or publicly available methods.
In this case, the Tribunal found that GOL’s director had the requisite expertise to be regarded as a “competent professional” capable of evaluating the project’s technological challenges. GOL’s AI-based system integrated multiple data sources, minimised false positives, and introduced a novel approach through bespoke algorithms and a substantial code repository. Such work was deemed an appreciable improvement rather than a routine adaptation of existing software. Consequently, the Tribunal held that the activities qualified as R&D, allowing GOL’s appeal against HMRC’s refusal.
Comment
In this case, Chat GPT significantly aided GOL by instantly translating foreign-language data into English. For instance, Chat GPT translated crucial documents from French into English, enabling GOL’s AI system to swiftly identify key financial risks and potential compliance issues that would otherwise have been difficult to detect. GOL integrated Chat GPT’s translations directly into their onboarding system, demonstrating the flexibility and efficiency of AI in enhancing real-time data analysis. One can easily see the benefits of such a system.
That said, and although it’s interesting to see a case where Chat GPT is given some positive judicial consideration, there may be broader concerns with any heavy reliance on AI. AI does still pose risks, including translation errors, inherent biases, and transparency issues, which may need to be carefully monitored. For that reason, I think, at least for now, there must be caution to ensure that human oversight remains.
