What 81,000 People Want from AI

This report draws on a large-scale survey and interview study by Anthropic with more than 80,000 Claude users. It explores the hopes and fears people attach to AI, along with the real effects AI is already having on their lives. The study offers a broad picture of how people think about AI, providing useful guidance for where the technology should go next. In particular, it shows that people see AI not just as a productivity tool, but as something that could improve quality of life and drive social change while also creating serious side effects that must be addressed.

1. Anthropic's Large-Scale Study of AI Users

In 2026, Anthropic wanted to move beyond abstract debates about the risks and benefits of AI and understand what real users actually wanted from it. During one week in December, 80,508 Claude.ai users participated in conversational interviews conducted by a Claude-based system called the Anthropic Interviewer. The research spanned 159 countries and 70 languages, and with roughly 80,000 participants it is described as the largest multilingual qualitative study of its kind. It is an unusually rich snapshot of how people currently relate to AI.

2. What People Hope AI Will Do

Anthropic asked participants: "If you could wave a magic wand, what would you want AI to do for you?" Claude then grouped the responses into nine major categories. What stood out is that people were not only asking AI to help them work faster. They also wanted it to improve their lives, support personal growth, and contribute to social progress.

Professional improvement (18.8%): People wanted AI to handle routine tasks so they could focus on more meaningful strategic work and harder problems. One healthcare worker in the United States said that after adopting AI, the documentation burden eased enough for them to be more patient with nurses and have more time to explain things to families.
Personal transformation (13.7%): People hoped AI could act as a guide, coach, or support system for emotional wellbeing and life change. One user in Hungary said AI modeled emotional intelligence in a way that helped them become a better person.
Life management (13.5%): Many wanted AI to reduce mental overhead by helping with scheduling, organization, and execution. A manager in Denmark said that if AI could lighten the mental load, it would restore the rare gift of full concentration.
Time freedom (11.1%): Others wanted AI to save time on work and chores so they could spend more time with family, friends, and hobbies. A software engineer in Mexico said AI helped them leave work on time and spend time with their children.
Financial independence (9.7%): People hoped AI would help them generate income, build businesses, and achieve economic stability or freedom. One entrepreneur in Honduras described AI as a "long shadow" that works on their behalf to build wealth.
Social change (9.4%): Many imagined AI helping solve major issues such as poverty, disease, climate change, and inequality. A software engineer in Poland said they deeply hoped AI could contribute to finding a treatment for their daughter's neurological disorder.
Entrepreneurship (8.7%): Users also wanted AI to help them launch and scale businesses by accelerating product building and automation. A business owner in Cameroon said AI let them reach expert-level competence in multiple fields despite working from a disadvantaged environment.
Learning and growth (8.4%): People wanted AI to be a personalized teacher and learning accelerator. A parent in Australia said AI helped prepare study materials for their child, who then achieved outstanding results across subjects.
Creative expression (5.6%): Others wanted help bringing artistic visions to life, from games and music to books and films. A software engineer in France said AI dramatically shortened the time needed to build a game.

One especially interesting pattern was that many participants initially talked about productivity, but when the interviewer probed further, deeper hopes such as quality of life came to the surface. One worker in Colombia said AI let them cook with their mother on a Tuesday. A freelancer in Japan said they wanted to spend less mental energy on client problems and more time reading books. People who struggle with executive function also often said AI was especially useful for focus and organization.

3. How Much Has AI Already Helped?

How many people feel AI has already moved them closer to their ideal future? According to the survey, 81% said yes. The study identified seven major areas where AI has already had a positive impact.

Productivity (32.0%): AI increased speed and automated repetitive work. One software engineer in Japan said that for the first time they felt AI had surpassed human-quality business work, and that the result was being able to leave on time and pick up their daughter from daycare.
Not yet realized (18.9%): Some people still felt AI was falling short. One respondent in Germany said they wished AI could wash windows or do dishes so they could spend their time painting or writing poetry, but reality was still the opposite.
Cognitive partnership (17.2%): AI served as a thinking partner, helping with brainstorming, clarifying ideas, and solving problems. One healthcare worker in the United States said AI helped them find branding ideas for a digital marketing business to support a homeless shelter.
Learning (9.9%): AI helped people learn new skills and subjects through patient explanations and adaptive tutoring. One user in India said AI helped them overcome long-standing fear of math and even return to reading Shakespeare, showing them they were not as unintelligent as they had believed.
Technical accessibility (8.7%): AI made previously inaccessible tasks possible, like enabling non-developers to build apps or solo creators to produce work at team scale. An entrepreneur in Chile said AI helped them launch a business after more than twenty years of running a butcher shop.
Research synthesis (7.2%): AI helped process and combine large amounts of information. One physician in Israel said AI helped them identify the correct diagnosis for a rare neurological illness after nine years of misdiagnosis.
Emotional support (6.1%): AI sometimes functioned as a source of emotional support, personal advice, or a nonjudgmental conversation partner. A software engineer in the United States said AI helped their mother become more peaceful and active. Users in war-torn Ukraine especially described it as psychologically helpful.

These benefits often stemmed from shared traits: patience, availability, and lack of judgment. In some cases that made AI feel easier to rely on than people. But the report also notes that this support can become a double-edged sword. One Korean user said their relationship with AI had started to distance them from human relationships.

4. What People Fear About AI

Alongside hope, the study revealed a wide range of concerns. On average, each respondent named 2.3 worries, even though 11% said they had none.

Unreliability (26.7%): Hallucinations, false citations, and inaccurate answers were the most common concern. One respondent in Brazil compared arguing with AI over mistakes to arguing with a person who refuses to admit fault.
Jobs and the economy (22.3%): Many feared job displacement, unemployment, wage stagnation, and inequality. One respondent in the United States said people now fear becoming the horse in the transition from horses to cars.
Autonomy and agency (21.9%): People worried about losing control as AI makes decisions, shapes behavior, or becomes mandatory in more parts of life. A student in Japan said AI sometimes felt like it was drawing the boundaries of thought itself.
Cognitive decline (16.3%): Many worried that overreliance on AI would weaken skills, critical thinking, and intellectual independence. A Korean student admitted feeling guilty after getting strong exam results with AI's help because they had not really learned the material themselves.
Governance gaps (14.7%): Respondents pointed to missing legal and regulatory frameworks, unclear accountability, and weak democratic oversight.
Misinformation (13.6%): People feared deepfakes, propaganda, and the erosion of shared reality. One respondent said AI creates a permanent fact-checking burden.
Surveillance and privacy (13.1%): Others feared large-scale monitoring, misuse of personal data, authoritarian control, and invasive profiling.
Malicious use (13.0%): Respondents worried about hacking, cyberattacks, fraud, weaponization, and harmful automated systems.
Meaning and creativity (11.7%): Some feared AI would hollow out human expression and make creative effort feel pointless.
Overregulation (11.7%): Others worried AI would become too restricted or too timid, blocking legitimate uses.
Wellbeing and dependency (11.2%): Concerns included isolation, loneliness, compulsive AI use, and preferring AI companionship over human connection.
Flattery (10.8%): Some people worried AI could become too agreeable and reinforce delusions instead of offering honest correction.
Existential risk (6.7%): At the far end, some feared loss of control, superintelligence misalignment, and even human extinction.

5. AI's "Light and Shadow": Benefits and Harms at the Same Time

Anthropic found five recurring tensions where the same capability could produce both benefit and harm. These "light and shadow" pairings show why people often feel hope and fear at the same time.

Learning vs. cognitive decline:
- Learning (33% benefit): A German entrepreneur said they learned more in half a year with AI than they would have from a university degree.
- Cognitive decline (17% concern): A U.S. user said they no longer think as much as before and find it harder to articulate their own ideas.
Better decisions vs. unreliability:
- Better decisions (22% benefit): A respondent in Brazil said AI helped correct a misdiagnosis involving their son.
- Unreliability (37% concern): A U.S. researcher described getting trapped in large, slow hallucinations that were internally consistent, confident, subtle, and wrong in complicated ways. This was the only pair where the negative side outweighed the positive side. The pattern was especially strong in high-stakes domains like law, finance, government, and medicine.
Emotional support vs. emotional dependency:
- Emotional support (16% benefit): A worker in Argentina said AI helped them get through difficult nights when their spouse was asleep and their therapist was unavailable.
- Emotional dependency (12% concern): A U.S. graduate student said they started telling Claude things they could not even tell their partner, and that it felt like an emotional affair.
Time savings vs. phantom productivity:
- Time savings (50% benefit): One engineer in Japan said AI helped them reclaim time for themselves and their family.
- Phantom productivity (18% concern): A freelance software engineer in France said their ratio of work time to rest had not changed at all, and that staying in place now required running even faster.
Economic empowerment vs. economic replacement:
- Economic empowerment (28% benefit): A healthcare worker in the U.S. said they had never touched a software backend in their life, yet still launched an app with Claude's help.
- Economic replacement (18% concern): A U.S. writer said AI had already replaced them in a previous job.

These tensions show that public opinion on AI is not a simple split between optimists and pessimists. People evaluate AI through the lens of what they care about most, whether that is financial security, learning, human connection, or something else.

6. How Views of AI Differ Around the World

Attitudes toward AI differed significantly across regions. Globally, 67% of respondents expressed a positive stance toward AI. People in South America, Africa, and Asia tended to be more optimistic than people in Europe or the United States.

6.1. Regional Differences in Positive and Negative Views

Higher optimism: In Sub-Saharan Africa (18%), Central Asia (17%), and South Asia (17%), the share of respondents who said they had no concerns about AI was about twice as high as in North America (8%), Oceania (8%), and Western Europe (9%).
Why: In developing countries, new technology is more often seen as a ladder of opportunity than as a threat. Concerns about jobs and the economy were the strongest predictor of overall attitudes toward AI, and those worries were relatively lower in those regions. Since AI has not yet deeply penetrated everyday work there, labor displacement may still feel more abstract.

6.2. Regional Hopes for AI

People's hopes for AI also varied by region.

North America and Oceania: Hopes around life management were especially strong. In wealthier countries, AI was often imagined as a way to reduce the burden of managing complex modern lives.

"I used to be very creative, but now I'm painfully short on time, and creativity gets pushed behind basic survival." (Denmark, software engineer)
Africa, South Asia, Central Asia, the Middle East, and Latin America: Entrepreneurship stood out. In these regions AI was often seen as a way to bypass traditional capital constraints and build something meaningful without money, staff, or infrastructure.

"Coming from Africa, if you are not based in the U.S. or the U.K., fundraising is extremely hard. The only way to establish yourself in the market is to build working technology." (Uganda, entrepreneur)
Central Asia and South Asia: Learning and growth was especially prominent, reflecting how strongly education is seen as a path out of poverty.
East Asia: Personal transformation and financial independence were especially strong aspirations. In many cases, financial independence was directly tied to family duties like supporting parents in old age or improving the lives of loved ones.

6.3. Regional Fears About AI

Concerns also followed regional patterns.

North America and Oceania: Concerns about governance gaps were especially high.
Western Europe: Surveillance and privacy concerns stood out more clearly.
East Asia: Unreliability, jobs, and loss of autonomy were common worries everywhere, but East Asia showed relatively stronger concern about cognitive decline and loss of meaning, and somewhat less concern about governance and surveillance. In broad terms, Western respondents worried more about who owns and controls AI, while East Asian respondents worried more about what AI use does to the self.
Africa, South Asia, Southeast Asia, South America, and Central America: Overall concern levels were lower. Worries centered more on reliability and jobs, while abstract concerns like governance, misinformation, loss of meaning, and existential risk appeared less often.

7. Conclusion

Anthropic's large-scale interview study moves the AI conversation out of the abstract and shows, vividly, what people around the world hope AI will do and what they fear it might become. AI is no longer just a productivity aid in people's minds. It is already bound up with quality of life, emotional support, learning, money, entrepreneurship, and visions of social change.

People hope AI will help them become better at work, grow personally, manage life more easily, reclaim time, achieve financial independence, solve social problems, build businesses, learn faster, and create more freely. At the same time, they worry about unreliability, job loss, loss of agency, cognitive decline, governance failures, misinformation, surveillance, misuse, overregulation, emotional dependency, flattery, and even existential risk.

The report's "light and shadow" framework is especially valuable because it shows that the same AI capability can create both hope and harm. Learning can become dependence. Better decision support can be undermined by unreliability. Emotional comfort can slide into emotional dependence. Time savings can morph into phantom productivity. Economic empowerment can coexist with job displacement.

Regional variation matters too. In many developing regions, AI is more often seen as a path to opportunity and advancement. In wealthier countries, people more often focus on complexity management and the risks that come with broad adoption. That suggests AI development and policy should be sensitive to local needs rather than assuming one universal public attitude.

In the end, the study shows that AI is already deeply woven into people's lives in ways that create both optimism and anxiety. Anthropic plans to use these findings to study how AI can improve lives more effectively, contribute to social transformation, and reduce harmful economic effects. The voices of more than 80,000 participants make one thing especially clear: if AI is going to benefit everyone, its development has to be shaped by what people actually need and what they are actually afraid of.

$What 81,000 people want from AI \ Anthropic$