New $5 Million Grant to Test AI Chatbot Safety in Crisis

Anthropic is offering $5 million in grants, typically ranging from $500,000 to $1.5 million, to fund research teams evaluating the safety of AI chatbots when used by distressed individuals.

The initiative follows a June 2025 study of 4.5 million conversations showing 2.9% were emotional exchanges, with researchers finding that current models often provide location-inappropriate crisis resources.

Applicants must submit proposals by September 21, 2026, and all resulting evaluations must be published as open-source resources to help improve AI responses to self-harm and mental health crises.