Anthropic Releases Findings from Global User Study on AI Expectations and Concerns - MLQ.ai

The people who most believe in AI are the ones with the least access to it

Here is a finding that should complicate every confident prediction about how artificial intelligence will change work, economies, and society: the regions most optimistic about AI are the ones with the least AI infrastructure.

Anthropic published a study this week — the largest multilingual qualitative study of AI attitudes ever conducted, with 80,508 respondents across 159 countries and 70 languages — that found people in Sub-Saharan Africa and Asia express 10 to 12 percent lower rates of negative sentiment toward AI than people in Western Europe and North America. Respondents from Sub-Saharan Africa showed particularly high aspirations for entrepreneurship and financial independence through AI. The same gap appears when comparing Latin America and Asia against North America.

The finding cuts against the standard narrative about AI anxiety. The assumption in wealthy economies — where AI has been most visible, most discussed, and most present in headline-making disruption events — is that the technology inspires both awe and dread. The data suggests otherwise. The people who see AI as an equalizer are not the people who have the most reason to know what they are talking about.

What people actually want from AI

The study's most cited finding is the geographic optimism gap. But the underlying data is richer than a single statistic. Anthropic asked respondents to describe, in open-ended conversations with a Claude-powered interviewer, what they most wanted from AI. The answers were surprisingly consistent across income levels and geographies.

Professional excellence was the most common response globally — 18.8 percent of respondents wanted AI to handle routine work so they could focus on higher-value problems. This was not a developing-world phenomenon. Healthcare workers in the United States described using AI to handle documentation and free up time for patients. Software engineers in Japan said AI let them leave work on time and cook with their mothers. The productivity pitch, it turns out, translates everywhere.

But the second-order findings reveal the divergence. In Sub-Saharan Africa and Asia, significantly more respondents framed AI as a path to entrepreneurship and financial independence — escaping economic constraints that felt immovable without a technological lever. In wealthier economies, the framing was more often about convenience and marginal productivity improvement, not transformation.

"AI is the great equalizer," Marc Einstein, research director at Counterpoint Research, told CNBC. "In rural Indonesia or Brazil, people have access to the same AI as in the U.S. or Japan." That is the optimism. The counterpoint is that AI may amplify existing vulnerabilities — digital exclusion, algorithmic bias, dependence on external systems — precisely in the places that have the least infrastructure to manage those risks.

The gap between aspiration and delivery

When asked whether AI had moved toward their stated vision, 81 percent of respondents said yes. But the study's methodology has a significant limitation: all respondents were existing Claude users. Anthropic was transparent about this — the sample skews toward people who found enough value in AI to keep using it, and likely toward more positive visions than a general population sample would produce. Nearly half of all respondents originated from North America and Western Europe.

This matters for how to read the geographic optimism gap. The people in Sub-Saharan Africa and Asia who responded are not a random sample of their populations — they are the AI-connected middle class, the people with enough education, infrastructure, and access to have a Claude account and use it regularly. For them, AI is already functioning as an equalizer in ways that it is not yet functioning for the average worker in Germany or Ohio.

The concerns tell a similar story. Across all regions, worries about reliability and accuracy were the most common — 22.3 percent of respondents named job displacement as their primary concern, spread fairly evenly across job categories. The voluntary-sample limitation means the people most worried about AI displacement — those who have been most exposed to it, or most displaced by it — are probably underrepresented.

The independent worker dividend

One of the study's most striking findings: independent workers — including entrepreneurs, small business owners, and gig workers — experienced more than triple the rates of economic empowerment from AI usage compared to salaried employees. The implication is that AI is particularly good at amplifying individual productivity in ways that translate directly to income, rather than increasing the output of people who are already employed in organizations.

Counterpoint Research's Einstein frames this as the agentic AI question. As models become capable of handling more complex tasks with less supervision — the trajectory Anthropic's February Cowork launch illustrated, spooking software and research stocks — the question is not whether AI will change professional work. It is whether that change will primarily benefit the people who can frame themselves as independent operators, or whether organizational workers will find their own leverage.

The geographic optimism gap may ultimately be less about where AI is going than about who feels entitled to claim its upside. The people who most believe in AI are, disproportionately, the people who have found a way to make it work for them — and that group, so far, looks less like Silicon Valley and more like Cameroon, Honduras, and India.