A house burning in the night, and a voice asking "can AI be trusted?". That is how the film Anthropic, the company behind the Claude assistant, released on July 9 opens. Uncomfortable questions follow, "who's gonna hit the brakes, if we need to?", "why do we have to have this stuff?", before the tone turns hopeful. It is called "There's hope in hard questions", and the questions were not written by creatives, they came from 52,000 Americans the company surveyed and 81,000 Claude users across 159 countries, interviewed, irony aside, by an AI tool built for the job.
At first glance this makes no sense, a company that lives exclusively off its product is paying for campaigns reminding you to doubt it, while the competition accelerates without asking permission. Look closer, though, and the paradox dissolves, leaving behind the most rational move this market has produced.
Official Anthropic film: "There's hope in hard questions".
This ad's customer is the skeptic
Anthropic does not live off the consumer racing to the cheapest model, it lives off the companies and developers who, according to industry estimates, account for roughly 80% of its revenue, including banks, public administrations and regulated sectors. That buyer is not seduced by hype, it buys predictability, compliance and trust. An ad that says "keep thinking" is speaking directly to the segment that has not bought yet precisely because it does not trust.
And there is a classic precedent: in November 2011, on Black Friday itself, Patagonia ran a full-page ad in The New York Times saying "Don't Buy This Jacket", and sales grew by around 30% in the months that followed. Volvo spent decades selling the fear of a crash. When the whole market shouts speed, the adult-in-the-room position sits vacant, and it is scarce. Anthropic was born in exactly that position, founded by former OpenAI staff who left over disagreements about model safety, and the film is the commercial continuation of that identity.
Asking first means choosing the stage
There is a second, finer layer: the fear the film invokes did not need to be created, because four decades of cinema, from The Terminator onward, planted the question "what if the machine turns against us?" in everyone's head. Anthropic does not create the doubt, it domesticates it, gives it beautiful cinematography, music and a hopeful ending.
And whoever asks first chooses the stage. "Who's gonna hit the brakes, if we need to?" spoken in the company's own film is a question with a controlled answer; the same question asked in a parliamentary inquiry is something else entirely. With AI regulation taking shape in Europe and the United States, the company that "invites the hard questions" today earns a seat at the table when the answers get written into law.
The uncomfortable part is that none of this requires cynicism, the mission can be sincere and the campaign calculated at the same time, and it is precisely that alignment between conviction and commercial interest that makes the move so good.
How to tell decorative doubt from the real thing
The real test begins when the film ends, and it applies to Anthropic as much as to any brand selling you humility:
- Look for the record, not the ad. The campaign points to a page where the company promises to show progress on answering the public's questions. Today, that page collects questions; the verifiable progress is not there yet. Come back in a few months and compare.
- Ask who verifies, because self-assessment is not an audit: serious commitments have independent third parties confirming them, deadlines, and consequences when they fail.
- Follow the cost. A promise only carries weight when it costs its maker something: turning down a contract, delaying a launch, publishing a failure. Without cost, it is a script.
- Apply the same test to your bank, your carrier, your social networks. "We take your privacy seriously" is the cheap version of this film.
The ad gets the essential thing right, hard questions really are the way. But the film's own central question remains unanswered outside of it, because for now, the one hitting the brakes is the one driving. Trust is not announced, it is verified.
#StaySafe
🙏🖖
