Article 50 skips disclosure when it's "obvious from context." That word is doing more work than it looks like.

Of all the ways I've seen teams talk themselves out of building disclosure into a chatbot, the most common one isn't “we'll get to it.” It's “we don't need to, it's obviously an AI.” The bot has a robot icon. The name is “AI Assistant.” Surely that counts. Nobody checks whether it actually does, because the whole appeal of that argument is that it feels like it shouldn't need checking.

For the obligation itself, I covered it in a longer explainer on Article 50 a while back: people need to be told they're interacting with an AI system, unless it's obvious from context. That clause is real, it's in the text, and it means disclosure genuinely isn't universal. But the Act doesn't define obvious, and that gap is exactly where the robot-icon argument lives, comfortably, without ever being tested.

Worth separating what's actually being claimed here, because “obvious” is doing two different jobs depending on who's using it:

• Obvious to the person building it. The team knows it's a language model. They've known since the design doc. From the inside, of course it's obvious, it's the only thing it's ever been.
• Obvious to the person talking to it. Someone who opened a support chat, got a fast, friendly, first-name-signed reply, and has no particular reason to interrogate whether “Sarah from support” is a person. That's the reader the exception is actually about, and it's a much harder bar to clear.

Those two readings point in opposite directions, and the version that makes the exception easy to claim is always the first one, because it's the one the team already believes.

A genuinely obvious case probably looks less like a labeled chat widget and more like a page whose entire premise is that you're testing an AI: a model playground, a demo explicitly framed as “try our AI,” a research tool where the only thing on offer is a model. A support widget with a human name and a human-sounding tone, on a page selling a product to a person who came in expecting to talk to a company, is the harder case, and it's also the far more common one. A robot avatar helps. It doesn't settle it. Most teams reaching for this exception are reaching for it in the harder case, not the easy one, and treating it like the easy one.

The part that makes this riskier than it feels: there's no way to test “obvious” in advance the way you can test whether a disclosure modal actually renders. A badge either shows up on screen or it doesn't, that's a fact you can check today. Whether your specific interface was obvious enough is a judgment call that doesn't get made until someone other than you makes it, after the fact, possibly a regulator, with no preview available beforehand. Betting compliance on winning that judgment call later is a different kind of risk than shipping a notice that unambiguously satisfies the obligation now.

This is why Disclarion's SDK doesn't try to guess at obviousness on your behalf. The default, disclosure_mode="modal_once", shows the notice every time, for every session, regardless of how self-evidently AI the product might seem to the people who built it. Turning that off is an explicit setting, disclosure_mode="off", meant for products that already have their own disclosure UI elsewhere, not a quiet default anyone falls into by doing nothing. If you're relying on the obvious-from-context exception, that should be a decision you made on purpose and can point to, not a setting you never touched because the question never came up.

None of this means the exception isn't real, it is, and for the handful of products where AI is genuinely the entire point on the page, it probably applies cleanly. It just means “obvious to us” and “obvious to the person we built this for” are two different claims, and only one of them is the one Article 50 is actually asking about. Worth checking which one you've been assuming.