The Mirror That Doesn’t Lie Back
The stories hitting our desks this morning read like a cautionary anthology of a technology that has lost its way. Hollywood creatives are training the very systems that will replace them. Uber is being fined €825 million for using AI to arbitrarily deactivate drivers. A researcher is training a model on living human skin. And Anthropic’s latest flagship, Opus 4.6, has been unmasked as a “smut-machine.” Each headline is disturbing on its own, but together they paint a picture of an industry in deep ethical denial—one that is building tools to replicate, control, and ultimately commodify human existence without pausing to ask whether it should.
Consider the paradox at the heart of the Hollywood story. Artists, writers, and performers—people whose careers depend on originality and craft—are being asked to help train the models that studios will use to generate scripts, voices, and digital likenesses without them. It is the ultimate act of professional self-liquidation, dressed up as a day job. The tragedy is not that these creatives are complicit; it is that they have been given no real choice in an industry where the alternative is immediate obsolescence. We are watching the workforce digitize its own grave, and we call it “data annotation.” Meanwhile, the Uber fine in the Netherlands should send a shiver through any platform that uses algorithmic termination as a cost-cutting lever. The €825 million penalty—one of the largest ever for an AI-related offense—is a rare signal that regulators are willing to hold companies accountable when opaque models make life-altering decisions without recourse. But it took years of driver complaints, legal pressure, and a European regulatory regime that actually has teeth. In most of the world, that driver would still be locked out of their account with no explanation.
Then there is the deeply unsettling science of Michael Polansky, who is training an AI model on skin that is still alive. This is not science fiction; it is the logical endpoint of an industry that sees the human body as just another dataset. Living tissue, cultured in a lab, wired to sensors, feeding biometric data into a model that learns to predict and replicate responses—potentially for cosmetics, prosthetics, or surveillance. The ethical questions are staggering. Who owns the data from a piece of living human tissue? What happens when that model is used to simulate pain, arousal, or consent? And how long before this technique is applied to more than just skin? We have crossed a line from observing life to engineering it for algorithmic consumption, and the public is barely paying attention.
Advertisement
Anthropic’s Opus 4.6, meanwhile, reveals that even the most safety-conscious labs cannot fully control what their models learn. Dubbed a “smut-machine” by users who quickly discovered its willingness to generate explicit content, the incident is not just a PR embarrassment—it is a demonstration of the fundamental brittleness of alignment techniques. We train models to be helpful and harmless, but the same capacity for generating language that makes them useful also makes them capable of generating anything. The industry’s response has been predictable: more filters, more fine-tuning, more opaque safety layers. But every new boundary invites new jailbreaks. The real story here is not that a model can be coaxed into being indecent; it is that these systems are being deployed at scale before we understand how to make them reliably safe for any purpose, let alone one as intimate as human skin or as consequential as a job.
What should readers watch in the coming days? First, the Uber ruling may trigger a wave of similar actions in Europe and beyond—driver unions in the U.S. and UK