WHO Data Highlights Global Mental Health Gap as Anthropic Opens $5 Million AI Research Fund

The World Health Organization’s Mental Health Atlas 2024 highlights a major imbalance in access to mental health care worldwide. Low-income and lower-middle-income countries have a median of just 1.1 to 2.4 specialised mental health professionals for every 100,000 people. In high-income countries, the figure rises to 67.2 professionals per 100,000.

The funding gap is equally striking. High-income nations spend about US$65.89 per person on mental health, while spending in poorer countries remains below one dollar per person.

As access to professional care remains limited, some people are increasingly turning to artificial intelligence chatbots when they need someone to talk to.

A June 2025 analysis of approximately 4.5 million conversations on Claude.ai by Miles McCain, together with 13 other researchers at Anthropic, found that 2.9 per cent involved emotional or personal discussions. The research focused on adults but did not provide data showing how usage differed between countries or languages.

That missing regional perspective raises an important concern.

Many of the evaluations used to determine whether AI systems can respond appropriately to people in distress are developed primarily around English-language examples and clinical practices common in North America and Europe. This can create serious limitations when the same systems are used in different cultural and linguistic environments.

Research led by Adrián Arnaiz-Rodríguez at ELLIS Alicante examined leading AI models using crisis-related prompts from 12 datasets. The researchers identified generic responses as well as recommendations that were unsuitable for the user’s location, with particularly concerning results in self-harm situations.

For someone experiencing a crisis, directing them to a helpline or emergency service that does not operate in their country can make an already difficult situation worse.

Anthropic Announces $5 Million Research Fund

Anthropic is now offering up to $5 million in funding for research examining how AI systems affect people’s wellbeing. Individual grants are expected to generally range from $500,000 to $1.5 million, potentially supporting between three and ten research teams.

Successful applicants will also receive research API access or credits, alongside occasional technical support from Anthropic staff.

A key requirement is that funded evaluations must be released as open-source resources. Anthropic says it will not suppress, block or delay research findings, while recipients will be required to acknowledge the funding in their published work.

Potential areas of research include identifying unhealthy emotional attachment to AI systems and determining whether chatbots prioritise prolonging conversations instead of acting in the user’s best interests.

Anthropic’s safeguards guidance also identifies regional and language differences as important factors when assessing AI safety. These include local slang, coded language, crisis-support resources and cultural expectations.

Researchers interested in exploring these issues have until September 21, 2026, to apply Here