Anthropic’s Hard Questions Ad Backfires Into an Anti-AI Warning

Anthropic’s newest ad for Claude opens on a burning house at night. It cuts next to facial recognition scanning a crowd, a homeless man on a sidewalk, laborers in a mine shaft and rows of tombstones arranged like a military cemetery.

The film, titled “There’s Hope in Hard Questions,” was supposed to prove that Claude’s maker takes AI’s dangers seriously. What it produced looked, to critics across tech and marketing circles, like the best argument against AI made all year.

A Burning House Opens the Pitch

The spot comes from the agency Mother and was directed by Myles McAuliffe. It landed on July 9, timed to a World Cup broadcast of Argentina against Switzerland, as the newest entry in Anthropic’s “Keep Thinking” platform, which launched in September 2025.

The first half plays like a trailer for something you would not want to watch alone. It moves through:

• A house engulfed in flames at night
• A crowd scanned by facial recognition software
• A homeless man sleeping on a city sidewalk
• Laborers working a dark mine shaft
• Rows of tombstones arranged like a military cemetery

Over those images, a voiceover asks, “Can AI be trusted?” and “Who’s gonna hit the brakes if we need to?” Only in the second half does the tone brighten, cutting to people embracing and someone watching the ocean, closing on softer questions like “Could AI help people stop feeling misunderstood?” The personal stories and questions Anthropic published alongside the film include a substitute teacher asking whether AI could speed up medical diagnosis. Since the ad went live, it has pulled in 3.3 million views on X alone.

The Cemetery Shot That Broke the Internet

One image did more damage than the rest combined. As the voiceover asks who is going to hit the brakes, the film cuts to rows of white headstones that multiple viewers said closely resembled Arlington National Cemetery.

Tech workers and marketing critics flooded social platforms calling the pairing tone deaf. Reactions ranged from confused to furious.

Anthropic is quite an amazing company. With the worst corporate communications ever.

That line, aimed at the campaign, was one of the comments TechCrunch pulled from the pile after the ad went live. Another poster, reacting to the same cemetery frame, called it exceptionally weird and sinister. One person distilled the whole backlash into a single question that has followed Anthropic since: can Anthropic be trusted?

Altman Piles On, Again

Sam Altman got there first. The OpenAI chief executive posted on X on July 14 that he assumed the ad was satire.

“i thought this was satire, kept looking for the handle to be spelled c1audeai or something,” Altman wrote in a post that quickly racked up attention. He did not stop there, going on to suggest that hard questions are welcome at Anthropic only for users the company deems worthy, a jab tied to accusations that Claude quietly downgrades some users’ access to its models.

This is not a new fight. Altman had already vowed OpenAI would never run ads that way back in February, after Anthropic’s Super Bowl spots mocked ChatGPT’s move into advertising. The two founders have sparred publicly since Altman and Dario Amodei’s team split ways over AI safety disagreements years ago. The timing now carries extra weight: both companies are reportedly preparing for public listings and confidentially filed IPO paperwork last month.

Built on 120,000 Hard Conversations

Anthropic insists the film was not improvised. It grew out of the Anthropic Public Record, a survey that asked 52,000 people in the United States to name their biggest hopes and fears about AI.

Layered on top of that is a separate round of interviews: 81,000 Claude users across 159 countries and 70 languages, gathered through a tool the company calls Anthropic Interviewer. Put together, Anthropic says more than 120,000 people worldwide fed into the research behind the film, a figure one trade outlet initially misreported as 12,000 before issuing a correction.

The company has also promised to keep taking public questions and report back on what it does with them, including where it comes up short. That consumer-facing goodwill push echoes other recent moves, like extending free access to Claude Fable 5 right before a planned price increase.

Anthropic’s Own Record Complicates the Message

The irony cuts deeper than tone. Futurism reported that Anthropic walked back a safety pledge in February, one that had promised to halt training a system if it could not guarantee proper guardrails were in place.

The same outlet reported that even as Anthropic publicly resisted Pentagon plans to use its technology for mass surveillance and autonomous weapons, Claude had reportedly been used to help select strike targets in Iran. An ad built on images of surveillance and mass graves lands differently against that backdrop.

There is a similar friction in how Anthropic spends money. Days before the ad’s backlash peaked, the company was finalizing a 20-year, $19 billion data center lease with bitcoin miner TeraWulf in Kentucky, the kind of industrial buildout that sits uneasily beside a commercial about laborers toiling in a mine.

Does the Dystopian Pitch Even Work?

Marketing analysts are not united on whether the discomfort was a mistake or the point. Sources disagree sharply on what the ad actually accomplished.

• Critics and tech workers argue the dystopian tone undermines the trust Anthropic is trying to build, especially the Arlington-style cemetery shot paired with talk of hitting the brakes.
• Brand strategists at FourWeekMBA frame it as a deliberate structural bet, arguing Anthropic is trying to own an emotional category before rivals can define it, since converging model capabilities leave brand as one of the few remaining moats.
• Fast Company’s analysis takes a third view: that a short commercial is simply the wrong format for a message about civilizational risk, regardless of how Anthropic had cut it.

What is not in dispute is that Claude’s underlying product keeps winning where the ad stumbled. A JetBrains developer survey from April found 46% of developers picked Claude as their favorite coding tool, against 9% for GitHub Copilot despite Microsoft’s distribution advantage, and Cursor, built on Claude, reportedly hit $2 billion in annualized revenue. That gap between product strength and brand execution is part of why competitors keep circling: undercutting Anthropic and OpenAI on coding pricing is exactly the kind of pressure a wobbling brand campaign cannot afford. Altman has separately claimed more Texans use ChatGPT for free than the total number of people who use Claude in the United States, a reminder of how far behind Anthropic still sits in raw consumer reach.

A Grand Prix Winner Turned Punchline

This is the third act of the same campaign in less than a year, and the reception has swung wildly each time.

The February spots worked because they were specific and funny, poking at a real product decision rather than civilization itself. Anthropic had pledged Claude conversations would stay ad-free and built a campaign around that single, provable claim. Hard Questions swapped a specific joke for an existential mood board, and the mood board did not land the same way.

None of this is happening on a shoestring. Anthropic is reportedly worth $350 billion following a $30 billion fundraising round in February, and the company is currently paying up to $400,000 a year to hire someone to run brand events and “large-scale brand activations,” a sign of how much weight it now puts on how the brand feels rather than just what the product does.

Anthropic has not issued a public statement addressing the backlash. The company’s only commitment on record is to keep publishing the hard questions people send in, and to report back on what it does about them.