It’s a fair question. Regulations aren’t written for machines - sometimes it feels like they’re not written even for humans!

But here is the thing, GenAI has come a long way and one of the things it is extremely good at is parsing and interpreting text - the reason being is because it has an awful lot of it to learn from. With the right infrastructure, domain fine-tuning, legal ontologies, and prompt engineering, today’s models can reliably interpret nuance, cross-references, carve-outs, and multi-layered regulatory logic. They don’t just keyword-match. They reason.

We’ve built Cardamon’s AI agents to operate across any regulation, any language, anywhere in the world. But we also layer in human validation where it counts. In this game, accuracy isn’t optional.

Are Regulators On Board?

Regulators differ from country to country and have different appetites. Overwhelmingly, we are seeing them embrace it and even want to use it themselves. I think the key nuance is that they are less concerned about how you get to the answer as long as you do get to the answer. If using GenAI gets you there quickly and reliably, then I don’t see any issues.

Essentially, while no one expects regulators to rubber-stamp black-box AI, we’re seeing growing support for explainable, transparent compliance tooling that enhances governance, not bypasses it.

Look at the FCA’s Digital Sandbox, MAS’s Project NovA!, or the SEC’s increasing focus on RegTech. The message is clear: regulators want to see firms harness tech to do compliance better, not just faster.