You'd assume any machine doing important work has two basic things: proof it was safety-tested, and an off switch. Researchers at MIT checked 30 of the most common 'autonomous AI' tools — the kind that go off and do tasks on their own — and found that most disclose nothing about safety testing, and many have no documented way to shut them down at all.
Here's why that matters: these aren't lab experiments. They're already deployed, doing real work. And if one starts misbehaving, there may be no clear, agreed-upon way to stop it — because the basic protocols simply don't exist.
So how does it touch you? More and more of the services you rely on — at your bank, your doctor's office, your job — are quietly handing tasks to these tools. If one goes haywire while handling your money or your records, the uncomfortable truth is that nobody may know where the off switch is.