Microsoft just gave us the perfect name for AI thieves: Kyle

Ketan Joshi reads Microsoft’s 30-second Copilot advertisement as an unusually candid picture of generative AI’s appeal: Kyle, a jealous office worker, sees his Black colleague Daniel praised for a forecasting model, then uses Copilot to analyze Daniel’s work, improve its charts, and make his own version “beautiful.” The ad ends with Kyle receiving the admiration Daniel had earned. Joshi argues that the comedy exposes a real power dynamic: AI lets people appropriate others’ labor and claim its results as their own, especially when the original contributors are less powerful or already overlooked. The ad’s racial and gendered implications, and its portrayal of Kyle’s envy turning into satisfaction, make the theft seem less like an accidental misuse than a product feature.

Joshi extends the criticism beyond AI training data to direct copying: users can feed existing work into a model and request a changed version that may be harder to detect or challenge. He says model outputs remain mediocre, requiring users to polish or disguise what they have taken. As an example, he describes a cleantech website republishing a degraded rewrite of his own work, leaving him with no meaningful remedy. He argues that platforms and employers could punish this conduct but have little incentive to do so, since AI hype creates permission for appropriation that benefits them. The central charge is that generative AI both relies on others’ labor and encourages users to misrepresent that labor as their own.