The CNET reporter arrived at AI on the Lot this year with a credential few in the room shared: a SAG-AFTRA membership card from the 2023 strike, when the union walked out over, among other things, non-consensual generative AI use of performers' likenesses. Aaron Pruner covers home entertainment and streaming TV, and he came to the Culver City gathering near the Amazon MGM backlot intending to test which read of generative AI matched reality, the deep fear or the heightened optimism that has split the industry for three years. He went wanting to change his mind. What he left with, by his own account, was a room running on product hype and techno-optimism, where concerns about human replacement and environmental damage were rarely mentioned, and where the means to bridge the gap from fear to understanding are lacking.
That gap is the story. AI on the Lot, billed by organizers as the world's biggest conference focused on AI in entertainment and media, drew roughly 2,500 attendees in 2026, per the CNET conference dispatch. It sits a short drive from studios that have spent the last year quietly wiring generative tools into previsualization, storyboarding, VFX, dubbing, and marketing. The conference is, in effect, the trade show for that wiring. So when the people building it cannot name what the wiring displaces, the silence is itself a data point.
Two films anchor the public version of the moment. Hell Grind premiered at Cannes and is cited as a working example of AI-assisted feature production. Dream of Violets screened at Tribeca as the first full-length AI-made film, per Pruner's reporting. Both are real artifacts with real festival credentials. Neither resolves the questions the conference is supposed to be asking, because neither was produced under the consent and labor framework the industry just spent two years negotiating.
That framework is not ancient history. The 2023 SAG-AFTRA strike ran 118 days, and the eventual contract introduced consent and compensation guardrails specifically for digital replicas and generative AI use of performers. The strike ended because actors agreed, grudgingly, that a regulated path was better than an unregulated one. A working actor who struck on those terms walking into the industry's flagship AI conference 18 months later and finding the regulated path barely a topic of conversation is not a personal anecdote. It is a measurement of how far the dominant AI conversation has drifted from the deal that made it legal.
What gets airtime in the room, per the CNET report, is the next model release, the next studio partnership, the next cost-per-minute number. The recurring absence is a structured answer to the obvious questions: who gave permission for the training data, who consented to likeness use, and who is paying for the compute. Each is solvable in principle. None of them is being solved at the conference where the people who could solve them are physically present.
The reframing is overdue. "Can AI break or remake Hollywood?" is the wrong question because it concedes the industry has only two possible futures, both determined by the technology rather than the choices made about it. The more useful question, and the one Pruner's conference visit quietly supplies, is which voices and what tradeoffs are structurally absent from the conversation the people inside the room are having. The 2,500-person answer is: a lot of them.
That answer is not a verdict. It is a checklist for the next time a press release describes a generative video tool as "transforming" the business, and for the next time a panel claims the technology is now "indistinguishable" from traditional work. The test is the one the strike set: did the people whose work, likeness, or environment the model touches agree to it, and on what terms. If the conference of record will not ask that question on stage, the work moves to the unions that wrote the rule, the regulators who enforce it, and the audiences who can refuse to watch when the answer is no.