In the broadest sense, AI Governance is the framework we use to ethically implement AI across business and society. From an enterprise-specific lens, AI governance is the framework used to control the unique business and regulatory risks associated with AI adoption. This is an area of focus with generative AI due to its agentic characteristics and the non-deterministic nature of its output.
We marvel at the creativity of prompting services such as ChatGPT. Part of that creativity can lead to unanticipated output. As AI gains more access to workflows, functions, and even takes over the development of programs, we need to govern the range of its output with guardrails and systems in place to ensure that creativity doesn't pose problems.
This is an especially acute issue in regulated industries such as financial services, government, and health care, where risk management is in many ways existential. Issues surrounding privacy and risk come to the regulatory fore.
As an example, let's take an AI function that is being widely adopted as of September 2024: AI-powered customer service interfaces. Many companies are looking to AI to help improve the customer service experience, which traditionally relied on call centers (which demand time and patience from the customer) or kludgy online navigation choices (that can be restrictive and frustrating).
Generative AI can tap into its agentic flexibility and creativity in addressing customers' questions on the fly. However, to do so, it often needs to learn from actual user inputs and data. Ensuring that private information isn't cross-contaminated or misappropriated is key, especially in a non-deterministic system. Service interactions must also limit hallucination and any potential risk to the consumer. Imagine a customer service request for a pair of skis that the customer suspects was damaged upon shipment. A smooth-talking AI may suggest the ski is perfectly fine, when in fact it poses a danger to the consumer.
AI Governance in many ways addresses the societal issues that are central to our identities as citizens of countries or geographic regions. As a result, there are national and even supranational agencies addressing AI Governance (such as WEF or the European Union).
Each is tailored toward societal norms that fit within the umbrella of each entity's supervisory purview. But the question of ultimately authority and rulemaking often comes into play. Companies will have to learn how to navigate this landscape in much the same way they do other forms of diffuse, global regulation. Some of the largest enterprises (such as Google and OpenAI) have developed their own frameworks in accordance with governments for tackling AI Governance.
World Economic Forum AI Governance Alliance
Effective Ventures - Centre for the Governance of AI
Google - Perspectives on Issues in AI Governance (PDF)
IBM: Artificial intelligence (AI) governance refers to the guardrails that ensure AI tools and systems are and remain safe and ethical.
AWS: "AI governance enables alignment with organizational goals, and ensures that AI technologies are ethically used and effectively managed. To that end, AI governance frameworks create consistent practices in the organization to address organizational risks, ethical deployment, data quality and usage, and even regulatory compliance, as well as managing the different cost patterns of AI workloads."