AI is a dangerous tool, but we keep using it because it is useful. If it were merely dangerous we would walk away from it. Instead it is in our phones and our operating systems, in HR departments and insurance companies, and it got there faster than any tool before it.
Some of that danger is the familiar kind. A chainsaw is dangerous too, but much less so if you learn how it works and practice using it. Much of what AI can do to you, to your privacy, your money, your attention or your judgment, can be handled the same way, with knowledge and good habits. The trouble is that the tools arrived before the lessons. Humans have had no time to catch up.
There is a second kind of danger that no amount of practice can solve. The people building these systems say, on the record, that AI will become more capable than we are, and that they do not yet know how to control it. You cannot train your way out of that. It sounds like science fiction, or horror, so we naturally want to dismiss it. It moves faster than anyone can keep up with, so we fall increasingly behind. We can measure what these systems do while we are watching. We cannot measure what they do when we are not. And we certainly cannot measure what they are capable of unless we can observe it.
Both kinds of danger come down to trust. Every time you use AI, and every time it is used on you, with or without your knowledge, someone is deciding how much to trust it with your data, your money, your security and your freedom to decide for yourself. The biggest of those decisions are being made by a small number of people, without you. Trust should be earned. Right now it is being decided for you.
Dangerous Robot exists to bring the danger into plain sight, with evidence you can check: what it does to you, what it does to society, what its makers promise, and what they have already failed to control. With that in hand you can choose better where you have a choice, and push harder where you do not. Choices add up to market pressure. Voices add up to policy. Both still work, but only while the choices are still ours to make.
This site is built with the help of AI. Every claim on it has a source, and a person who stands behind it.