It's true that the EU AI Act includes transparency mandates for general-purpose AI, but calling it always present feels a bit strong to me, especially with the drawdown risks for smaller players.
These rules aren't always a fail-safe if a company decides the compliance burden is too high, creating an exit option.
For instance, if a start-up in Montreal finds the Article 53 requirements too complex or expensive, they might just decide not to offer their AI model in EU countries at all.
That means the mandate isn't truly universal, but instead is conditional on continued market participation.
It limits the downside for businesses, but it also limits the reach of the regulation.