The SKF ad arrives with a ready-made justification: Garbo really did appear in a 1921 promotional film for the company when she was fifteen and still called Greta Gustafsson. That historical footnote is the fig leaf. What it's covering is a 99-second demonstration that AI image synthesis can produce something that looks vaguely like a screen legend while retaining absolutely none of what made her one. The problem isn't technical fidelity — the figure on screen is clearly lit like a movie star, dressed well, shot with care. The problem is what the technology cannot access. Garbo's screen presence was built from contradiction: icy hauteur and latent vulnerability, comic deadpan wrapped in operatic sensuality. These are not visual properties. They're the residue of a particular consciousness navigating performance, and no amount of image synthesis produces them. What SKF has instead is a Stepford robo-star, as the Guardian puts it — the classiest possible presenter of a TV shopping channel. The brand logic is almost admirably literal. SKF manufactures ball-bearings. Ball-bearings are smooth. Garbo was famously smooth (apparently). Therefore: Garbo for ball-bearings. The ad even has AI Garbo remark, 'One of my first films was for SKF and now I'm back for my last.' The word 'last' is doing a lot of work there — a stilted joke that lands nowhere near the deadpan comedy Garbo actually deployed in films like Ninotchka. The ownership question is the most structurally interesting part of this story, and the Guardian is right to flag it. By reportedly drawing from images and pictures outside Garbo's film catalogue rather than from the MGM archive, SKF may have sidestepped the studios — who hold copyright over her on-screen performances as IP — and dealt instead with distant relatives who proved more amenable. The article wonders aloud whether Humphrey Bogart, Gary Cooper, and Rita Hayworth have similarly incentivizable descendants. That's not a rhetorical flourish. That's a business model being sketched. What the ad actually reveals is the gap between what AI can replicate and what art actually is. Garbo's impact, as the piece notes, rested on something other than smoothness. The qualities that made her iconic — that particular quality of seeming both present and unreachable — are precisely what no synthesis can reconstruct from images, because they weren't in the images to begin with. They were in the encounter between a specific person and the camera at a specific moment in time. The AI version is a simulacrum of the aesthetic surface without the generative tension underneath it. The ethical dimension is real but perhaps less novel than it feels. Publicity rights, likeness rights, and posthumous image licensing have existed for decades. What's new is the cost curve: generating a passable AI likeness is now cheap enough that it becomes commercially rational to approach distant relatives of dead stars. The guardrails the Guardian calls for would need to address not just the technology but the economic incentive structure that makes this kind of thing worth doing.
