This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
“We use AI responsibly” has mostly been a values statement. In Europe, it is about to become something that can be checked against a legal text.
On Friday, the EU’s Artificial Intelligence Act was published in the Official Journal as Regulation (EU) 2024/1689, seven weeks after the Council gave it final approval. It enters into force twenty days after publication, on August 1. The obligations then arrive in phases: bans on prohibited practices from February 2025, rules for general-purpose AI models from August 2025, most of the rest from August 2026, and some high-risk product rules a year after that. The top fines, for prohibited practices, reach €35 million or 7% of worldwide annual turnover, whichever is higher.
Nobody has been fined under the Act, and nobody will be for some time. Codes of practice and much of the guidance are still to be written. So what follows is a prediction, and it comes with that caution attached.
Responsible-AI language becomes testable
Most corporate AI principles were written for a world without a shared yardstick. The AI Act supplies one, with defined categories and obligations. I’d expect journalists, analysts, procurement teams and advocacy groups to start reading company claims against it. A polished “trustworthy AI” page that doesn’t match how a company describes the same systems to European authorities is exactly the kind of gap that turns into a story.
Legal vocabulary will read like a verdict
The Act sorts AI uses by risk. “High-risk” is a legal classification tied to particular uses, such as hiring or creditworthiness decisions. It is not a finding that a product is dangerous. That distinction is unlikely to survive a headline or a search snippet. To most readers, “Company X’s high-risk AI system” will sound like an accusation.
Communications teams should know now which of their products may fall into which category, and have a plain-language explanation ready before someone else supplies one. Search results tend to keep the first framing that gets traction.
The reputation splits by jurisdiction
Global companies have usually told one AI story everywhere. That gets harder when the same feature launches in one market and is delayed in another. A preview has already arrived from a different law. Last month Apple said Apple Intelligence and some other features probably won’t reach EU users this year, citing the Digital Markets Act. That came less than two weeks after it announced ChatGPT inside Siri for later this year.
As the AI Act phases in, I’d expect more “available everywhere except Europe” announcements. Each will invite the same question from the press: what would the European version have had to disclose?
There is a more ordinary transparency piece too. The Act requires that people be told when they are interacting with an AI system, and that certain AI-generated or manipulated content, such as deepfakes, be disclosed. For brands using chatbots in customer service or synthetic imagery in marketing, those notices become part of the visible product in Europe. Screenshots of missing ones will travel.
What to do with a prediction
None of this calls for a crisis plan today. It does suggest a few quiet moves. Inventory where the company uses AI in ways customers can see. Check that the public AI principles match the compliance reality. Decide who answers when a reporter asks how a product is classified.
My guess is that the companies that look most credible when the main obligations apply will be the ones whose public statements in 2024 were modest enough to be true.