The inappropriate use of AI tools can lead to serious breaches of client confidentiality. Carrie Langridge, Head of General, Compliance and Tax Investigations at ICAEW, and Sarah Brook, Director of Investigations, outline how risks arise, what firms can do to mitigate them, and the potential implications of getting it wrong.
Key takeaways
- Accountants’ confidentiality duties still apply when using AI.
- Entering client data into public AI tools can risk unauthorised disclosure.
- Firms should regularly review AI policies and train staff on confidentiality risks.
- AI policy breaches are not automatically reportable to ICAEW; potential Code of Ethics breaches require assessment.
Confidentiality is one of the five fundamental principles at the heart of ICAEW’s Code of Ethics. As firms increasingly use AI tools, it is vital that their policies, procedures and staff training address risks to client confidentiality. This will help firms protect client information and comply with the Code’s principles and wider data protection requirements.
“We’re seeing some instances where firms or members are entering confidential information into open source AI tools without appreciating that the information they’ve entered is then available to a much wider group of people than the Code of Ethics’ confidentiality requirements would allow,” explains Carrie.
As well as having implications for professional conduct and data protection compliance, sharing client information, even inadvertently, can damage firms’ business relationships and reputation. “If a firm is responsible for a data breach because it either lacks sufficient policies, or isn’t enforcing them, that could be very damaging given the type of data accountants hold about their clients,” emphasises Carrie.
Why and how are AI risks to confidentiality arising
“Many firms have good policies and procedures about how staff should use AI tools, and some firms have their own AI platforms, which keep all data within the firm,” says Sarah.
“But there are still some firms who might not fully recognise where the information they enter into generative AI platforms is going and how it might be used.” Even where firms have their own ring-fenced AI systems and specific policies, there is still the risk of staff using public AI tools or AI features within software in ways that don’t align with firm policy.
“People sometimes don’t appreciate or understand the implications of what they're doing,” explains Carrie. “And that might be simply because we’re all still learning how to use AI and, as we’re learning, AI is continually developing, offering up new tools and encouraging people to use them.”
“When you’re entering data into an AI tool,” says Sarah, “it can feel as though the information stays within your work system or on your laptop. It might not feel like posting publicly online or in a chat forum. But, in reality, if the tool isn’t within a secure environment, entering information into it can be akin to posting it publicly.”
As an example, an accountant looking into tax implications for a client might enter relevant information – such as the client’s name and that they’re looking to sell some of their shares – into an open source AI platform to determine how the tax treatment might work.
“What they aren’t realising is they’ve now put some very confidential client information out into the open market,” stresses Carrie. “It’s pretty much like putting it on a virtual notice board; other people now have access to that sensitive information.”
What should firms be doing to address AI confidentiality risks?
Firms should familiarise themselves with the ICAEW Code of Ethics, particularly subsection 114 on the principle of confidentiality, and ensure their internal AI policies, procedures and training reflect the Code’s requirements.
R114.1 in the Code of Ethics states that: “A professional accountant shall comply with the principle of confidentiality, which requires an accountant to respect the confidentiality of information acquired in the course of professional and business relationships.”
“We expect firms to have policies addressing the use of generative AI, including the potential risks to confidentiality,” says Sarah. Although firms can’t necessarily control every individual action, they can show they've taken reasonable steps to help reduce the risk by setting clear policies, communicating them effectively and implementing training. The pace at which AI is evolving also makes regularly reviewing policies essential to keep up with the latest available tools.
The first bullet in R114.1 says an accountant shall “be alert to the possibility of inadvertent disclosure, including in a social environment…”. In the past, professional training focused on reminding people about confidentiality in terms of having a conversation or sharing a document. For example, don’t sit in a train carriage and talk about a client’s tax affairs, or leave a document out on the bar at after-work drinks.
“A social environment has traditionally been the train or pub scenario, but we all know it can now be the online environment as well,” explains Sarah. “We’ve learned how to deal with confidentiality issues around social media; and now, it's time to recognise that putting the information into an open source AI platform is essentially the same thing. AI is the latest inadvertent disclosure risk people need to be aware of.”
When do firms need to report AI confidentiality breaches to ICAEW?
If a firm identifies that someone has put confidential information into open source AI, or has breached its AI policy, it needs to consider all the implications, including its duty to report.
“When deciding on the actions you need to take, focus on the individual case,” advises Sarah. “Think about what data has been released, whether it was accidental or intentional, and what the potential consequences might be for the client and the firm. Consider how to mitigate the breach, as well as any associated reporting obligations, whether to ICAEW or under data protection rules.”
In deciding whether to report a breach to ICAEW, firms should approach the issue in the same way they would any other conduct issue. The key question would be whether the action potentially breached compliance with the principle of confidentiality as laid down in the Code of Ethics. “That’s going to come down to the detail of what’s been shared,” says Carrie. “Was the breach inadvertent or intentional? How did it happen? Was the client’s name used? Did you input project names? Did you use information that would allow somebody to check back and identify the entity?”.
“It’s not necessarily the case that if you identify that someone has used AI against firm policy, then it should be reported to us,” she stresses. “Firms have policies on all sorts of things, and unless an internal policy breach also involves an identifiable breach of the Code of Ethics, it isn’t reportable to ICAEW.”
“However, if, after investigating, you decide there might have been a breach of the Code, you should be mindful of the duty to report misconduct,” says Sarah. “Firms also need to consider their wider obligations to inform their clients of any confidentiality breach, and the need to comply with data protection legislation, including reporting requirements under UK GDPR.”
Using AI responsibly and wisely
Firms need to manage their approach to deploying AI as they would any other area of business risk, and this includes recognising potential risks to client confidentiality from inappropriate or unauthorised use of AI tools and systems.
“If you want to use AI, then use it wisely and responsibly,” stresses Carrie. “Think carefully about the information you’re putting into it, and whether you are happy for that to become more widely available. Think not just about what you’re getting out of AI, but what you’re putting into it - and the consequences that could bring.”