Roel Willems advises supervisory boards, executive teams and data and AI leaders on data and AI strategy, governance and regulation. Twenty years in data and AI, eleven of them at Ahold Delhaize and Albert Heijn.
A proposal promised compliance and never explained how. That absence is the argument. Compliance is not a property of a system, it is a property of a system inside a particular organization, and no vendor can encode your promises. Here is the one question that tests the claim.
A loyalty migration made data more accurate and much harder to use. The only thing standing between the numbers and a bad decision was someone who remembered the change. That safeguard was never in a system, and it's exactly the context your AI can't inherit.
Boards heard "delay" and moved AI governance down the priority list. The fact remains that EU AI Act obligations arriving first did not move at all. And the list of systems those rules touch? Producing it is not the hard part. Keeping it true is.
On 29 June 2026, the Supreme Court's Slaughter ruling cracked the legal basis for EU-US data transfers. It's not a verdict, it's a warning with lead time. Why "your data stays in the EU" and "we don't store it" answer a narrower question than the one you're actually asking.
A senior colleague found a model eighty percent better than anything she'd used all year. Better at what, though? And in your own work, how would you know if the next version quietly lost it? You work through the same chat box daily with almost no view of what the model behind it can do.
A senior colleague found the best model she'd used all year. A day later, an export directive switched it off for everyone. She went back to Opus and kept working. But what happens when the model that vanishes isn't helping a person, and is instead running a process the business can't switch off?