Anthropic launches $5 million initiative to study ai's impact on wellbeing

Anthropic has announced a $5 million grant program aimed at funding independent research into how artificial intelligence affects user wellbeing. The initiative will support researchers in developing open-source evaluations to assess the impact of AI models on those who use them. Grantees will receive direct funding, access to Anthropic's models, and technical assistance, while maintaining full independence and publishing their work as open-source projects. AI systems have become integral to how people work, learn, and solve problems, and they also serve as conversational partners and sources of emotional support. However, the industry is still developing clear standards for how models should behave in sensitive interactions, such as when users seek companionship or use AI to navigate mental health crises. Wellbeing evaluation is particularly complex, as it requires contextual understanding rather than simple assessments of accuracy. For example, a user in distress may not immediately disclose self-harm intentions, and a response that seems appropriate in one context could be harmful in another. A response that might be reasonable for weight loss advice could be inappropriate for someone with a history of disordered eating. Anthropic has implemented safeguards to identify such conversations and ensure appropriate responses, and it publishes research on user interactions with its models to inform the development and evaluation of these safeguards. The company emphasizes that the right approach must evolve alongside AI models and their applications. The grant program seeks evaluations that are rigorous, transparent, and capable of informing future safeguards and benchmarks. Anthropic is also sharing guidance from its Safeguards team on what constitutes a robust wellbeing evaluation, along with common challenges that can limit its effectiveness. Applications for the grant are due by September 21, with selected applicants notified by October 5. The company also provided additional updates, including improvements to Claude Fable 5’s biology safeguards that reduce false positives and minimize fallbacks. Mariano-Florentino (Tino) Cuéllar has been appointed as Anthropic’s first Chief Global Affairs Officer.