John Margerison on
AI risk and governance
What boards and leadership teams need in place before AI goes wide: an acceptable use policy, a plan for shadow AI, and a clear view of who is liable when AI, or an AI vendor, gets it wrong.
5 newsletters, 2 videos, 6 in the media. Updated 16 Nov 2026
John's approach
John's take
John's warning to boards is that AI liability does not sit safely with the supplier. When an AI system makes a decision that harms a customer, an employee or a job applicant, the organisation that deployed it is in the frame. Boards that assume otherwise are, in his words, about to become unstuck.
He argues that three things need to be in place before AI goes wide. An acceptable use policy that says plainly which tools staff may use, with what data and for what. A plan for shadow AI, the tools people are already using without approval. And a clear line of accountability for every AI system in use, so that someone in the business owns each one and can explain what it does.
Why he sees it this way
The risk is already here. John has written that shadow AI, staff using unapproved tools with company data, is one of the fastest-growing security gaps in business, and that most boards have not clocked what it means. Court cases such as Mobley v. Workday, in which an AI hiring tool's decisions were challenged, show that responsibility for AI decisions is being tested in practice, not just in theory.
He also follows how fast the technology outruns the rules. He has argued that regulators should review AI law at least monthly, because frontier capabilities are moving faster than any annual review can keep up with. Until the rules catch up, governance inside the business is the protection that counts.
What he's watching
- Liability cases. How courts in the US, UK and EU decide who is responsible when an AI system or an AI vendor gets it wrong.
- The EU AI Act. Its rules phasing in between 2025 and 2028, and what they ask of businesses that deploy AI, not only those that build it.
- UK AI legislation. Calls for a dedicated UK AI bill to address risks to human rights, and whether one follows.
- Shadow AI and security. Whether unapproved AI tools drive the rise in cyber attacks he has warned about, and how companies bring that use into the open.
- Insurance. How insurers word policies for AI-related claims, and whether boards check their cover before an incident rather than after.
John's work on this
Everything →Questions about AI risk and governance
What is shadow AI?
What are the risks of shadow AI?
What should an AI acceptable use policy include?
Can you sue a company for using AI?
Does liability insurance cover AI?
On John's mind
On AI regulation
"AI is developing faster than regulation can keep up. More frequent changes in legislation are needed to keep AI safe, and safely accessible, for all."
John Margerison, Founder and CEO, YesBusiness AI16 Oct 2026Get in touch
For press, speaking or working with John.