This piece argues for an AI Social Contract as a safeguard, with staged licensing and human oversight in “gray zones.” It suggests that imperfection itself should be a design principle. Do we need governance frameworks like this — or is existing liability law enough?