Anthropic set out to prove it takes the dangers of AI seriously. A large part of the internet decided it had made a horror trailer instead.

The company’s brand film, “There’s hope in hard questions,” opens on a house engulfed in flames, then moves through crowd surveillance, a homeless person asleep on the street and rows of cemetery headstones, while a voiceover asks whether AI can be trusted and “who’s going to hit the brakes if we need to.” Released on 9 July and aired during a World Cup match, it was built to show a safety-first company willing to sit with the public’s fears. Within days it had been mocked across X and YouTube, most pointedly by OpenAI chief executive Sam Altman.

The reaction matters beyond the rivalry. Trust-and-safety messaging is the most contested ground in AI marketing, and Anthropic has staked its brand on being the responsible alternative to faster-moving labs. The episode is a rare, high-profile case of responsible-AI creative doing the opposite of its intent — amplifying dread rather than confidence — and it lands while marketing teams across the sector are being asked to write their own reassuring narratives about the same technology.

What did Anthropic’s ad actually show?

Anthropic’s film pairs a run of unsettling images with unanswered questions before turning hopeful. As the visuals shift from disaster and surveillance to families, healthcare and nature, the questions soften too, from “can AI be trusted?” to “could AI help people stop feeling misunderstood?” It closes on the lines “There’s hope in hard questions” and “Keep thinking,” the tagline of the broader brand platform Anthropic launched in September 2025 as its first major campaign for Claude. The film was made by the agency Mother.

The campaign is not a stunt, and that is part of what makes the reception striking. Anthropic says the work draws on its “Public Record” research: a survey of 52,000 Americans, responses from 81,000 Claude users across 159 countries and 70 languages, and dozens of focus groups, with the voices in the film sourced from a US roadshow earlier in the year. The company frames the initiative as an obligation of its public-benefit structure — “people have a lot of hard questions about AI,” its announcement reads, and “it’s our job to address them.” Viewers are invited to submit their own questions at claude.com/hard-questions, and Anthropic has pledged to report publicly on what it does in response, including where it falls short.

Anthropic’s problem was tone, not honesty

Anthropic’s misstep was execution, not the decision to acknowledge risk. The single image critics returned to was a field of headstones — read by many as Arlington National Cemetery — placed under the question of who will “hit the brakes” on AI. For a company asking to be trusted, pairing a warning about runaway technology with a military graveyard read less as candour than as a threat, and that juxtaposition is where the goodwill drained.

The mockery was swift and public. Sam Altman wrote on X on 14 July that he “thought this was satire,” joking that he kept checking whether the account posting the ad was a parody handle rather than Anthropic’s real one, before adding a sharper jab about the company restricting who gets access to its models. The most-liked comment on the film’s YouTube upload called it “dystopian marketing slop.” Writing in The Atlantic, Lila Shroff described the ad as “tone-deaf,” arguing it left viewers unsure whether Anthropic was warning them about AI or asking them to embrace it. A widely shared post on X put the strategic objection plainly: that the ad seemed to scare people about AI, then position its maker as the only party humble enough to be trusted.

Not all of the response was hostile. Some marketers read the film as a genuine creative risk — an attempt to move Anthropic from a feature claim, “safe,” to an emotional identity that rivals would find harder to copy. But the distinction that matters for the reader is between reaction and result: the backlash is well documented, while any effect on Anthropic’s sales, sign-ups or brand tracking is not, and no such figures have been released. What the episode proves is that the creative misfired with its audience, not that it damaged the business.

What the misfire shows about AI trust messaging

Anthropic’s experience points to a failure mode that marketers writing about AI can name and design around. Acknowledging a technology’s harms can build trust — brands own up to industry problems all the time — but only when the tone resolves into reassurance the brand can actually deliver. The film spent most of its length in fear and left its answer thin, so the dread is what viewers carried away. Owning a harm works on screen when the piece also shows the harm being addressed; leaving it hanging over a graveyard does the opposite.

Two practical cautions follow, and neither requires Anthropic’s research budget to apply. The first is that image and voiceover carry meaning together: a shot can attach an association to a line that the script never intended, and a cemetery beneath a question about stopping AI is exactly that kind of accident. The second is that doom-adjacent creative needs testing against a real audience before it ships, not only against the strategy that commissioned it — a room of people watching the cut cold would have flagged the graves in seconds. The lesson is not that a capable team got this wrong; a research-heavy, well-funded one did, which is precisely why the discipline of pressure-testing tone belongs in every marketing team’s process. It also sits inside a wider shift marketers are already navigating, in which brand safety is no longer about where an ad runs but what AI says about a company.

Anthropic has invited the public to submit their own hard questions and promised to publish what it does about them. Whether the campaign is remembered as a brave break from AI advertising’s conventions or as a cautionary tale about tone will depend less on the film than on what the company publishes next.