Anthropic asks the public to grade its answers to AI's hardest questions

TL;DR

Anthropic launched a public initiative on July 9 asking people to submit their hardest questions about AI—on jobs, safety, family, science, and power—and committed to publicly tracking the specific actions it takes in response, including where it falls short. It’s an unusual move: a frontier lab inviting open-ended outside scrutiny and promising a paper trail. Whether it becomes accountability or marketing depends entirely on the follow-through.

What happened

The announcement opens with the kind of questions the company says it keeps hearing: who decides the rules for AI, whether AI can give children a better future, whether it makes the world more dangerous, whether it can help scientists cure diseases. Anthropic’s framing is blunt—“People have a lot of hard questions about AI. It’s our job to address them.”

The initiative sits on a substantial pile of groundwork the company has already done. It cites the Anthropic Public Record, a survey whose first round asked 52,000 Americans about their biggest hopes and concerns; a separate survey of 81,000 Claude users across 159 countries and 70 languages, run through its automated Anthropic Interviewer; dozens of in-person focus groups and convened sessions with groups whose work bears on AI’s societal questions; and its Economic Index research, which studies real-world Claude usage through anonymized data.

Institutionally, the effort connects to the Anthropic Institute—an internal research effort aimed at the most significant challenges AI poses to society—and to the company’s Long-Term Benefit Trust, which provides oversight of its public-benefit mission. Anthropic is structured as a Public Benefit Corporation, and the announcement leans on that status: the company points to its safeguards work, alignment and interpretability research, free model access for scientists, and a fellowship pairing early-career Claude users with nonprofits as evidence it takes the mission seriously.

The new commitment is the reporting loop. Submit a hard question at the company’s hard-questions site, and Anthropic says it will “publicly track and report the specific actions we’re taking,” and be clear about the ways it might fall short of its stated goals.

Why it matters

AI companies usually publish the questions they already know how to answer. Inviting open-ended public scrutiny—and attaching future reporting to it—sets a more useful standard: whether company behavior changes, not whether the launch page sounds thoughtful. If Anthropic ships regular reports naming concrete actions and honest misses, that becomes a benchmark competitors will be pressured to match.

The difficult part begins after collection. Public questions are broad, conflicting, and rarely reducible to a product requirement. “What happens to my job?” does not translate cleanly into a roadmap item. The value of this initiative will depend on whether Anthropic connects the questions to specific decisions, evidence, owners, and deadlines—the boring machinery that separates a commitment from a sentiment.

The fine print

The announcement does not say how often Anthropic will report, in what format, or who inside the company owns the follow-through. There are no named metrics and no deadlines yet. The survey numbers are self-reported, and the research it cites is Anthropic’s own. None of that makes the effort hollow—but the mechanisms that would make it verifiable are still to be published.

The public has never been short of questions about AI. The experiment is whether a frontier lab can turn them into something more binding than a FAQ.