GenAI Needs Dedicated Governance. Here’s Why...and How
As generative AI (GenAI) becomes a mainstream enterprise technology, new risks are emerging. Companies must ensure its use is responsible, ethical, and carefully controlled.
Enterprise leaders have a lot to consider as use of GenAI evolves and scales. It’s not just about deriving value from the technology.
According to Deloitte’s latest report on the State of Generative AI in the Enterprise, “regulatory uncertainty and risk management have risen in organizations’ lists of concerns to address”.
Here at Amdocs, we’ve noted that this shift in emphasis is especially apparent in highly regulated sectors like financial services. With the EU’s new AI Act set to be applied from next year, we expect GenAI governance to command increased attention. However, while many enterprises see GenAI as an evolution of data analytics, bolting GenAI onto existing data governance frameworks won’t address the inherent risks. A more tailored approach is needed.
Navigating GenAI’s Risk Landscape
GenAI holds great potential for improving efficiency, accelerating innovation, and enhancing customer experiences, but it also brings new risks. These can range from the accidental misuse of data to unintentional bias or model hallucination. Some issues are intrinsically linked to the indeterministic nature of the large language models (LLMs) on which GenAI is based. For instance, LLMs may generate different responses to the same input due to their probabilistic sampling methods and the inherent randomness of the decoding process.
The associated risks for financial services GenAI use cases such as loan processing, insurance estimations, and anti-money laundering can be immense. Under the EU’s AI Act, companies using GenAI for high-risk applications like these face specific obligations surrounding factors including data quality, documentation, human oversight, and traceability. This is indicative of requirements that are likely to emerge in other parts of the world too.
In such a dynamic regulatory landscape, it’s no wonder enterprise leaders want to get on top of GenAI governance. But regulatory compliance is just one part of the equation. It’s also vital that GenAI is used in an ethical way to earn trust and ensure it delivers positive outcomes for customers, employees, and other stakeholders. This demands a focused approach rooted in a solid understanding of how GenAI works and how it can go wrong.
Benefits of a GenAI Governance Framework
A dedicated GenAI governance framework with robust policies and standards minimizes exposure to risk for the enterprise and its customers alike. Policies can encompass everything from broad principles of ethical AI to very specific matters such as the avoidance of bias.
For instance, a policy on AI ethics might cover principles of fairness, accountability, transparency, and non-discrimination. It could also include a commitment to human oversight and guidelines for ethical decision making. A policy on AI bias might look more specifically at bias detection methods and tools, strategies for bias mitigation, and guidelines on audits and fairness assessments.
As well as reducing risk and enabling the business to get ahead of regulations, policies and standards enforce considered, responsible use of GenAI throughout the enterprise. If industry can set a precedent for high ethical standards, rules imposed by authorities may not need to be heavy-handed. As Therese McCarthy Hockey from the Australian Prudential Regulation Authority (APRA) explained at the Australian Finance Industry Association Risk Summit 2024:“we are not adding to our rule book at the moment, [but] we will be using our strong supervision approach to stay close to entities as they innovate and consider management of AI risks.”
5 Pillars of Ethical AI
GenAI governance is a complex, multi-faceted and ever-shifting journey. The issues raised above are just a few examples of things that must be considered. So, we’ve created an enterprise playbook to help you on your way. It details five pillars of ethical AI which focus on topics like policies, compliance, risk management, roles and responsibilities, transparency, explainability and so on. The playbook also explains how the pillars can be used to form a strong foundation for an AI Center of Excellence that governs responsible and ethical use of the technology across the enterprise. To learn more about the 5 pillars, click below.
