Basil Yaqoob

As of August 2, 2026, the EU AI Act's high risk provisions are enforceable, meaning agentic AI systems now legally require human oversight and conformity assessments. Fines for non compliance can hit 15 million euros or 3% of global revenue.

What's interesting is that "human oversight" isn't clearly defined in most implementations I've seen. Does it mean a human reviews every action before it happens? Only high risk actions? A log someone can audit after the fact?

I've seen a few teams building agent products wrestle with this exact call for every action type, not as a compliance checkbox, but because it genuinely changes how much you can trust the thing day to day.

Working on a finance side AI tool, and one thing that surprised me: Stripe's invoice object doesn't have an "overdue" status at all. It only tells you "open" or "paid" overdue is something you have to derive yourself by checking if it's open and past its due date. If you build against the API assuming "overdue" is a real field, you'll confidently tell someone nothing is overdue when their business is full of unpaid invoices.

Feels like a small thing, but it's the kind of gap that's easy to miss until it actually costs someone money. Made me a lot more paranoid about assuming any third party API's status fields mean what they sound like they mean.

Curious if others building against Stripe/QuickBooks/accounting APIs have hit similar "the field name lies to you" surprises, feels like a common trap.