AI Acts: How Can Companies build Their AI Governance Frameworks?

Originally published on May 25 2024 in Towards AI

There has been a flurry of excitement surrounding the adoption of the EU AI Act last week which is being widely heralded as the world’s first regulatory act for AI (see Europe sets benchmark for rest of the world with landmark AI laws (Reuters, May 22, 2024), World’s first major law for artificial intelligence gets final EU green light (CNBC, May 21, 2024). Perhaps the reason for this interest lies in the potential economic impact of the Act. For example, McKinsey & Company estimates that just Generative AI (including apps such as ChatGPT or GitHub Copilot) could add the equivalent of $2.6 trillion to $4.4 trillion annually to business revenues — and over 75% of the value will be derived from embedding Generative AI for currently underutilized business functions such as customer operations, marketing and sales, software engineering, and R&D.

The EU AI Act is effectively a regulatory framework for legal enforcement of pledges regarding AI-related activities and behavior such as those signed in February by many of the world’s top tech companies and organizations during the 2024 Global Summit on Ethics and Artificial Intelligence. The UNESCO Statement signed by GSMA, INNIT, Lenovo Group, LG AI Research, Mastercard, Microsoft, Salesforce and Telefonica focused on ethical behavior in the development of and potential application of AI.

However, the EU AI Act has moved questions about AI governance from within the realm of tech companies to become an issue for every company. This means companies need an AI Governance Framework to govern their processes and ensure that any use of AI meets the basic criteria of being fair, transparent, accountable and respecting privacy. Most companies, particularly SMEs, don’t have the capacity or know-how to build an AI Governance Framework from scratch so this article aims to share some of the available resources to get started.

What is an AI Governance Framework?

First it is worth considering why an AI Governance Framework is important. Simply put, without a structure in place to govern potential issues such as algorithm bias, violations of privacy and perhaps the most-often cited concern, the use of AI for malicious purposes, companies may be exposed to risks (some preventable, others non-preventable).

A whitepaper by Securiti (March 3, 2024) explains that an AI Governance Framework ensures the responsible development and deployment of AI systems via a “structured set of regulations, policies, standards, and best practices intended to regulate and govern AI technologies’ development, application, and use. It serves as a guide to ensure AI systems are developed and utilized ethically, responsibly, and in accordance with legal standards.”

Getting started with developing a Governance Framework

A useful starting point to understand the latest key issues is the Global AI Ethics and Governance Observatory , a collaboration between UNESCO , the Alan Turing Institute , and the International Telecommunication Union (ITU) This aims to be “a centralized repository of information, insights, and resources aimed at guiding policymakers, regulators, academics, the private sector, and civil society in navigating the ethical and governance dimensions of AI.” Resources include which research, toolkits, and best practice-sharing “across a range of issues related to AI ethics, governance, responsible innovation, standards, institutional capacities, generative AI, and neurotechnologies” (UNESCO ).

Unlike most company processes which are proprietary, the topic of AI and the related ethical challenges are believed to be so critical to society that many companies have made their templates and learnings publicly available. For example, Rolls Royce has been developing The Altheia Framework since 2020, and has been open about its journey and made its template free online. From a corporate perspective this move has contributed to a view of the company as being transparent. AI Guardian offers a downloadable AI Ethics Policy Template . For a comprehensive list of toolkits and policy considerations, AI Ethicist.org has put together an A-Z list of Frameworks, guidelines and toolkits . A roundup of some of the best or most comprehensive AI policies and frameworks by country can be found in this list by MarkTech Post (May 2, 2024).

Key components of an AI Governance Framework include establishing guidelines based on fairness, transparency, accountability, and privacy (see my article Who should govern AI? regarding the challenge of defining the values for fairness). The Framework must cover issues such as data security including how data is handled; developer accountability and responsibility; bias minimization; and ongoing risk assessments. To achieve this, the Framework must be tailored and relevant to the needs of the company using it. For example, how much sensitive data is essential? What policies are needed regarding data management? What are the compliance requirements for the local culture, industry, business sector, the legal jurisdiction and the specific type of AI? How can staff be trained and educated? What structure is in place for oversight such as board structures, ethics committees and departmental processes?

Planning for a Global, Sustainable Future

AI transcends borders and is a global technology therefore it is important to create a framework that can incorporate international collaboration even for a small company. The framework provides a foundation for ensuring that AI technologies are developed and deployed in a manner that is ethical, responsible, and aligned with societal values. It isn’t about using buzzwords such as ethical and accountable but ensuring that companies are able to incorporate AI into their business to create efficiencies and develop in a sustainable manner.

The EU AI Act — Additional Reading

“The Artificial Intelligence Act” (Details of all the EU AI-related Acts) https://artificialintelligenceact.eu/the-act/

“A Guide to Mitigating AI Risks” Boston Consulting Group — November 29 2023 https://www.bcg.com/publications/2023/a-guide-to-mitigating-ai-risks

“AI and Product Safety Standards Under the EU AI Act” Carnegie Endowment for International Peace — March 5 2024 https://carnegieendowment.org/research/2024/03/ai-and-product-safety-standards-under-the-eu-ai-act/

“The EU AI Act: What it means for your business “ EY — March 15 2024 https://www.ey.com/en_ch/forensic-integrity-services/the-eu-ai-act-what-it-means-for-your-business