Anthropic opens $5m grant programme for AI wellbeing research

Anthropic opened applications for $5 million in grants funding independent research into how AI use affects the people who use it, open to clinicians, psychologists, methodologists and other specialists rather than restricted to AI researchers. The stated goal is a set of open-source evaluations and benchmarks the wider industry can use to measure effects on user wellbeing — a category Anthropic said is currently under-measured compared with capability and safety benchmarks.

Grantees will receive funding, model access and technical support, but Anthropic said they will design their own studies and work independently of the company’s direction, publishing whatever they find as open-source projects that any developer, including competitors, can use. Initial applications close on 21 September, with applicants notified by 5 October on whether they will be invited to submit fuller proposals.

The programme arrived in the same week as a separate pilot giving outside researchers access to aggregate Claude usage data, and the two shared a common argument: that data and research capacity relevant to AI’s real-world effects sit overwhelmingly inside a small number of labs, and that funding external, independently published work is one way to widen that base.