The legal considerations of AI for enterprises
AI is no longer a back-office technology project. It is now a boardroom issue that affects hiring, procurement, compliance, cyber risk and customer trust. For CEOs, the question is not just how quickly they can adopt AI, but whether their business is ready for the legal responsibility that comes with it.
CEOs now own the AI decision
AI used to be treated as a technology decision. That period is over. The CEO now has to understand where AI is being used, what decisions it is making, who checks the output, and what happens when something goes wrong.
Several business leaders are already making this shift visible. John Margerison of XFactorAI has warned that boards are underestimating the legal risk attached to third-party AI tools, while Satya Nadella is pushing Microsoft toward AI agents that change how work is done, and Klarna’s Sebastian Siemiatkowski has linked AI adoption directly to the shape of his future workforce. Microsoft’s 2026 Work Trend Index makes the same point in different language: the constraint is no longer what people can do with AI, but how work is structured around it.
That puts AI firmly on the CEO’s desk. It is no longer enough to ask whether the technology works. The better question is whether the business is set up to use it safely.
Vendor contracts will not protect you on their own
Many enterprises still treat AI like old software procurement. They buy a tool, sign the contract, pass the risk to the supplier, and move on. That is a dangerous habit.
AI tools do not just sit in the background. They rank candidates, assess customers, draft communications, recommend actions, and shape decisions. Once those systems affect people, money, access, or opportunity, the company using them is part of the risk.
This is where John Margerison’s warning matters. His view is that too many boards assume responsibility sits neatly with the vendor. He points to Mobley v. Workday, where companies using Workday’s AI hiring tool were also named as co-defendants, as an early sign that deployers can be pulled into legal claims even when the tool was built elsewhere. His proposed fix is simple: appoint one accountable owner for AI governance, audit the tools already in use, and revisit the assumptions inside vendor agreements before disputes arrive.
That should not be controversial. It is basic corporate hygiene. The problem is that many companies have moved faster on adoption than accountability.
Workforce change creates legal risk too
The loudest AI debate is still about jobs. That makes sense, but the legal question is wider than headcount. If AI changes how people are hired, managed, measured, promoted, or removed, it creates a paper trail that can later be challenged.
Klarna is one of the clearest examples. Its CEO has said the company’s workforce could fall from around 3,000 to 2,000 by 2030 as AI spreads through operations. Reports have also noted that Klarna had already reduced headcount sharply from its earlier size, partly through hiring restraint and automation.
That may prove commercially smart. It may also create hard questions. Were jobs removed because tasks vanished, or because management overestimated what AI could do? Were customers affected? Were staff consulted properly? Were protected groups hit harder than others?
The CEO does not need to block every workforce change. But they do need a clear record of why decisions were made, what evidence supported them, and how the company checked for unfair outcomes.
AI agents raise the stakes
The next legal problem is not just employees using chatbots. It is AI agents acting inside company systems.
Microsoft’s Build 2026 announcements show where enterprise software is heading. The company announced new AI initiatives including autonomous workplace assistants, AI agent devices, and a push to remake computing around AI rather than traditional apps.
That changes the risk profile. A chatbot gives an answer. An agent may take an action. It might schedule a meeting, update a record, draft a contract clause, trigger a workflow, or move information between systems. That is useful, but it also makes mistakes more expensive.
Enterprises need rules for what agents can and cannot do. They need approval points for sensitive actions. They need logs that show who authorised a task, what the agent did, and what data it touched. Without that, a company may struggle to explain a bad decision after the fact.
The CEO must treat AI as risk management, not just growth
AI can make companies faster. It can cut costs, improve service, and help teams do work that used to take far longer. But the CEO who treats it only as a growth story is missing half the issue.
Jamie Dimon has been unusually blunt about this. JPMorgan’s CEO has said AI will create areas where the bank can win, but also areas where competitors may take business from it, meaning the firm has to be “really, really good at it.” He has also linked AI to cyber risk, which he has described as one of the bank’s biggest concerns.
That is the right framing. AI is not a side project. It changes competition, cost, hiring, security, compliance, and customer trust at the same time. CEOs do not need to become technical experts. They do need to ask sharper questions.
Where is AI already being used? Which systems affect customers, employees, or regulated decisions? Who owns governance? What happens when a vendor tool fails? Which decisions need human review? What evidence would the company show a regulator, claimant, or journalist?
Those questions are not exciting. They are the work. The CEOs who ask them now will move faster later, because they will not have to rebuild governance in the middle of a crisis.