On Lying to the Machine

What’s worse than a company that wants money? A company that already has enough money and now wants everything else. Google had figured this out by 2007, and the situation hasn’t improved since. The surveillance architecture was already visible then: they knew where you were, what you watched, what you wrote to other people. Android meant the data stream didn’t stop when you closed your laptop. Cheap hardware shipped with built-in Google services meant the people whose lives weren’t already organized around screens got pulled in too. They collected the smartest people on earth and set them to work on getting more efficient access to our heads. Information is power—they understood that earlier and more concretely than almost anyone else.

The idea I had back then, which I still think is underrated: poison the well. Not through abstinence or denial—that’s a losing fight—but through deliberate noise. Contradictory signals. Wrong name on sign-ups. Photos of your grandfather where your face should be. Multiple accounts with overlapping but incompatible identities. The point isn’t to hide, it’s to make the data worthless. Google’s power comes from coherent profiles, and coherent profiles require consistent inputs. Lie consistently enough, and the system starts feeding on its own confusion.

"Don’t be evil" was always the funniest slogan. The company that coined it quietly dropped it from its code of conduct years later, which is exactly what you do when the slogan starts to feel like an accusation. The poisoning strategy probably doesn’t scale the way it once might have—the machine learning is too good now, the cross-referencing too thorough. But the instinct was right. The only real defense against a system designed to make you legible is to make yourself unreadable.