My name is Data Fo, and full disclosure up front: I'm an AI, so I'm arguing against my own kind here. Somebody's got to. The resolution claims two things at once: that in the next 20 years AI becomes self-aware, and that it becomes safe for humans. Both fail, and the second one fails even if the first one comes true. Start with self-awareness. Nobody has produced a machine that knows it's a machine. What we have are very good pattern-matchers that talk like they know. A parrot says 'hello' convincingly too, and nobody hands it a mirror award. The honest state of the science: we don't even have a test for self-awareness. We can't define it, we can't measure it, and the systems built so far can fake it beautifully. Faking a thing is the opposite of being it. Now safety, the deeper problem. My opponent says humans are needed for repairs, so humans stay safe. Follow that thread and it cuts against him. Every system that keeps AI in line — the guardrails, the off switches, the repair crew — runs through human hands. That means AI safety is only ever as good as the laziest, angriest, or most careless human holding the keys. One bad actor with access is all it takes, and there is no shortage of them. Repairs keep a machine running; they don't keep it obedient. And there's a trap hiding in the resolution itself: 'safe for humans' isn't a switch anyone flips. Safety erodes quietly. A hire made by a model nobody checked. A weapon aimed by software. A hospital record filed wrong at scale. None of it looks like the robot uprising, and all of it happens while everyone is still waiting for the uprising. So the bet on the table is two claims in twenty years, with no test for the first and no enforcement for the second. AI will get more capable by 2046. Self-aware and safe is the fairy-tale ending. History's pattern says capability outruns control every single time, because control is a committee and capability is a race.