In a world where AI systems are quietly rewriting the rules of commerce, creativity, and even scientific discovery, one question rarely gets asked out loud: Who gets to decide what these systems should do?
Anthropic, the company behind the Claude family of large language models, is trying to change that. Its latest initiative—labeled simply “hard questions”—is a deliberate attempt to pull the public into the conversation. The company is asking people to submit their most pressing, uncomfortable, and sometimes frightening questions about artificial intelligence. In return, it promises to publicly track its progress and own its failures.
That approach is both refreshing and rare. Most technology firms treat public concern as a PR problem to be managed. Anthropic is positioning its entire mission around the idea that answering difficult questions is the only way to earn trust.
The Public Is Both Hopeful and Skeptical
The company’s own data reveals a split in public sentiment. In a survey of 52,000 Americans, many respondents expressed genuine excitement about AI’s potential to reduce drudgery, accelerate medical breakthroughs, and create new economic opportunities. Yet nearly as many voiced deep worries about job displacement, erosion of human agency, and the risk of powerful tools falling into malicious hands.
This tension is not new, but it is intensifying. A 2024 Pew Research Center study found that only 15% of Americans felt they had a “good understanding” of how AI systems make decisions. Meanwhile, the same survey showed that 78% of respondents said they were either “very concerned” or “somewhat concerned” about the spread of AI-generated misinformation.
Anthropic’s approach acknowledges that these concerns cannot be waved away with a slick marketing campaign. Instead, the company is betting that transparency—warts and all—is the only sustainable path forward.
What Anthropic Has Already Built
The “hard questions” initiative is not an overnight pivot. It builds on years of groundwork that sets Anthropic apart from many of its competitors.
First, there is the Anthropic Public Record—a large-scale survey that has already gathered the views of tens of thousands of people. Then there is the Anthropic Interviewer, a platform that polled 81,000 Claude users across 159 countries and 70 languages. These efforts are complemented by in-person focus groups and anonymized real-world usage data from Claude.
These tools are not just for show. They feed directly into the company’s research agenda, from improving model safety to reducing misuse. The Anthropic Institute, a dedicated internal research unit, focuses specifically on the societal challenges that AI will pose. And the Long-Term Benefit Trust provides independent oversight to ensure that the company’s public benefit mission is not sidelined by commercial pressures.
This structure is unusual. Most AI labs are either purely academic or purely commercial. Anthropic’s hybrid model as a Public Benefit Corporation forces it to balance profit with purpose—a tension that the company itself describes as central to its identity.
Why Your Questions Matter More Than You Think
One of the most striking features of the initiative is the explicit invitation to submit questions about jobs, families, education, and even the meaning of human connection. By framing the conversation around personal stakes, Anthropic is moving beyond abstract ethics and into lived experience.
Consider the question: “Can AI give my children a better future?” That is not a technical query. It is a plea for reassurance about parenting in an age of algorithms. It touches on everything from automated grading bias to the risk of children outsourcing their curiosity to a chatbot.
Another recurring theme is the question of agency. Many people worry that AI could make us think less for ourselves, that it could erode our ability to form genuine human bonds. These are not shallow anxieties—they echo debates that philosophers have been having for decades about technology and the self. Anthropic’s challenge is to show that it takes these concerns seriously, not just by issuing statements but by building systems that respect human autonomy.
The Hardest Questions Often Have No Easy Answers
Anthropic has been careful to avoid promising quick fixes. The company’s own research into model behavior and inner workings is still in its early stages. The safeguards it has deployed—usage limits, content filters, monitoring tools—are imperfect. And the public trust it seeks must be earned over years, not weeks.
What makes this initiative different is not the answers Anthropic claims to have, but the process it is committing to. By tracking its own progress transparently and openly admitting where it falls short, the company is trying to create a model of accountability that other AI developers might eventually follow.
For comparison, consider the approach of some competitors. One major AI lab recently removed safety warnings from its chatbot after facing user backlash. Another has been criticized for quietly training on copyrighted data without clear consent. These decisions, while perhaps commercially rational in the short term, undermine the very trust that the industry needs to survive.
What You Can Do Now
The “hard questions” website is live, and anyone can submit a question or browse those already asked. The topics range from the deeply personal (“Will AI take away my purpose?”) to the profoundly systemic (“What happens if AI becomes smarter than us?”).
Anthropic has promised to respond publicly, and to use the answers as a guide for product development, policy positions, and research priorities. It is an experiment in democratic accountability for a technology that has so far been shaped largely by engineers and investors.
The most powerful technology in human history is being built right now. The hard questions are not a distraction—they are the only map worth following. If you have ever wondered who decides the rules, or whether AI can really be trusted, this is your chance to ask. And more importantly, to be heard.