AI governance principles for boards
B · Useful, limited KPMG and the INSEAD Corporate Governance Centre distilled from interviews with directors on over 25 boards; credible practitioner guidance, but principles only, with no survey data or tested outcomes.
Summary based on the full text·Original: about 35 minutes to read
In short
A 22-page report published in April 2026 by KPMG International with the INSEAD Corporate Governance Centre (Annet Aris, Theodoros Evgeniou). It proposes five principles for how boards should oversee AI without managing it: strategy, technology and security, workforce, trustworthy AI, and the board's own working methods. The principles were built from group discussions and interviews with experienced directors, investors and AI specialists across the Americas, Europe and Asia. Its central claim is that boards must govern at two speeds, short-term controls and long-term system-shaping choices.
Key numbers
- 5governance principles, broken into 22 sub-principles
- over 25company boards the steering committee members collectively serve on
- twenty-twofurther companies and organisations represented on the advisory council
As reported by the source.
English·PDF·opens assets.kpmg.com
What it reports
- Principle 1, strategic oversight: boards should balance short-term productivity pressure against sizeable early investment in data, infrastructure and change management whose returns are uncertain, and consider effects on business model, partnerships and M&A.
- Principle 2, technology and security: technology stack choices are a board topic because they set cost, speed and control. Boards should know which AI functions are strategically critical, review vendor dependency (including a vendor failing or being legally barred from supplying), and weigh sovereignty and microchip supply.
- Principle 3, workforce: boards ask management to define which decisions always need meaningful human oversight ("human-in-the-loop" and "human-on-the-loop"), and expect a workforce strategy that addresses whether more or fewer employees are needed, by department or skill group.
- The report flags entry-level jobs and apprenticeships as a talent pipeline risk boards should oversee.
- Principle 4, trustworthy AI: fairness, privacy across the supply chain, energy and water impact, and handling conflicting rules across jurisdictions.
- Principle 5, the board itself: update committee charters with AI responsibilities, adapt composition, adopt a legally vetted AI use policy for directors, and require reporting on where and why AI is adopted and whether it is a risk, an opportunity or both.
- Boards should treat AI-specific metrics, escalation routes, insurance and incident response as part of the risk framework.
How they got it
A steering committee of four senior directors and an advisory council of 13 named members were convened for group discussions and one-to-one interviews on real board dilemmas. Dilemmas were grouped into company, ecosystem and internal-board perspectives, prioritised, and turned into principles. No survey, sample statistics or outcome testing.
Limits
- Consensus guidance from senior directors; it does not show that any principle improves outcomes.
- KPMG sells advisory and assurance services on the same topics.
- Explicitly non-binding and "illustrative", so it gives few thresholds a board could audit against.
Why it matters for companies
For a company without board-level AI oversight, this is a ready agenda: which AI capabilities are strategically critical, which vendor dependencies are acceptable, and which decisions must stay with people. The workforce principle puts headcount direction, reskilling and entry-level roles on the board's list, which gives HR a formal route to raise them. Requiring management to report where AI is used and whether it is a risk or an opportunity is a low-cost first step. The report gives structure, not answers, so the company still needs its own risk appetite and metrics.
Summary based on the full text. Read the original for the full argument and data.Open PDF on assets.kpmg.com
- Tags
- Governance