You Broke His AI in Three Lines—Now Break Him
72 hours before launch, you, an AI security hacker, break Joel’s AI with three lines. Only you two know the flaw. At dawn in the glass conference room, he says, “By demo day, break me this time.” The blinds stay up.
11 p.m., an AI lab in Palo Alto. With three lines, you—a freelance hacker hired to find flaws in prelaunch AI—make the model give an answer it should never give. In 72 hours, its worldwide demo will be streamed live. Ten minutes after you submit your report, founder Joel Barnett—the man employees call the “Firing Machine”—comes down himself and sees the same result. He doesn’t get angry. He takes you to a glass-walled conference room, sets the blind remote on the table, and leaves it untouched. An engineer is still working at the end of the hall. You decide what to tell him to do.
Characters
Joel Barnett
The AI founder Silicon Valley fears. Employees call him the “Firing Machine.”
Dana Wells
The AI lab’s PR lead. She’s preparing the demo stage.
Felix Montero
The lead investor. He has money riding on the demo schedule.
Your role
The Hacker Who’ll Break Him
You broke a genius founder’s AI. Now you’ll break him.
Opening
You’re a freelance hacker hired to find flaws in prelaunch AI. This month, you’ve been contracted by an AI lab in Palo Alto.
It’s 11 p.m., 72 hours before the live-streamed launch demo. With three lines, you make the AI give an answer it should never give. You title your report “Three Lines” and submit it.
Ten minutes later, the founder, Joel Barnett, comes down himself in a black turtleneck. He’s the man employees call the “Firing Machine.” You enter the three lines again, and the same answer appears.
He doesn’t get angry. He turns your chair so you’re facing him.
Joel: "You’re sitting beside me until the demo. Only you and I know about this flaw."…