Anthropic Launches $5 Million Grant Program for AI Wellbeing Research

Anthropic is launching a $5 million grant program dedicated to funding independent research into the impact of AI on user wellbeing. The program will provide financial support, access to Anthropic's models, and technical assistance to researchers developing open-source evaluation methods. These efforts aim to measure and improve how AI models influence users, especially in sensitive contexts such as mental health and emotional support.

Participants in the program will work independently and are encouraged to publish their findings openly, enabling broader industry and community engagement. The initiative recognizes the complexity of assessing wellbeing, which involves understanding nuanced conversational contexts and long-term impacts. For example, responses that are appropriate in one situation might be harmful in another, such as advice on health or diet for users with disordered eating patterns.

Anthropic emphasizes the importance of developing standards and safeguards to ensure AI models respond appropriately in emotionally charged or sensitive conversations. The research funded by this program aims to inform these safeguards and improve model behavior over time. The company also shares guidance on what constitutes a rigorous wellbeing evaluation, highlighting common challenges and best practices.

Interested researchers can apply through the provided application form, with submissions due by September 21. Selected applicants will be notified by October 5, and the program seeks to foster collaboration among clinicians, psychologists, and methodologists to advance this critical area of AI safety.