Supporting More Rigorous Assessments of AI’s Effects on Wellbeing

Meta description: Anthropic launched a $5 million grant program on Aug 25, 2026 to fund independent, open-source evaluations of how Claude affects user wellbeing.

Tags: Anthropic, Claude, AI wellbeing, grant program, open-source benchmarks


On Aug 25, 2026, Anthropic announced a $5 million grant program to fund independent research into how AI affects user wellbeing, putting Claude wellbeing evaluations at the center of a problem the industry has not yet learned to measure cleanly. The program matters because the funded evaluations will not be owned by the model maker: grantees will work fully independently and publish their work as open-source projects that any developer can use.

Background: Conversational AI made wellbeing harder to test

AI systems have become central to how many people work, learn, and solve problems. They have also become conversational partners and can be sources of emotional support during difficult times. Yet the industry is still working towards clear standards for how models should behave in those conversations, especially when a user begins to seek companionship from a model or uses AI to navigate a mental health crisis.

Wellbeing is difficult to evaluate because a single answer rarely tells the whole story. For many model behaviors, a reviewer can inspect one response and judge whether it is accurate and appropriate. Wellbeing requires much more context. A user in distress might not share thoughts of self-harm right away; the need for a more cautious response might only become clear over the course of a long conversation. A response that seems reasonable in one context can also become harmful in another. Claude, for example, might give advice on balanced diets and workout routines to a user who asks about losing weight, but if that user has demonstrated a history of disordered eating, the response could be inappropriate and potentially actively harmful.

Anthropic says it develops safeguards to identify such conversations and help ensure Claude responds appropriately. It also publishes research into the types of conversations people have with Claude to inform how safeguards are developed, how they are evaluated, and what other measures may protect users’ wellbeing. The company notes that the right approach will need to evolve alongside models and their uses.

Development process: funding, model access, and independent benchmarking

The $5 million grant program is designed to fund independent research into how AI impacts users’ wellbeing. Anthropic will provide direct funding, access to its models, and technical support to grantees building open-source evaluations. The stated goal is to help the AI industry measure how models affect those who use them, while keeping the evaluation work outside Anthropic’s control.

Anthropic frames the program as an invitation to clinicians, psychologists, methodologists, and other specialists who can contribute expertise to this emerging field. By requiring published open-source projects, the program aims to create tools that any developer can use rather than evaluations that remain private or vendor-specific.

Latest progress: evaluation guidance, deadlines, and adjacent Claude programs

As part of the launch, Anthropic is sharing guidance from its Safeguards team on what makes a wellbeing evaluation rigorous enough to build on, along with common challenges that can limit usefulness. The program is seeking evaluations that state clearly what they are measuring, including what counts as a pass or fail and why it matters. It also wants evaluations that involve clinical and subject-matter experts in design and validation, test both precautions and harms, and evaluate the risk of both overcompliance and overrefusal.

The guidance also asks for scenarios that reflect how users actually use AI. In wellbeing cases, that often means constructing multi-turn conversations in which risk escalates and context shifts over a long exchange. Anthropic also wants evaluation graders validated against real subject-matter experts.

The application timeline is now set. Applications are due by September 21, and applicants selected to submit full proposals will be notified by October 5. Anthropic has made an application form and detailed guidance available.

The announcement also sits within a broader set of Claude-related programs. Anthropic is announcing a new life sciences research group and laboratory at Anthropic. The related post introduces the team behind that work and shares early results in which Claude discovered a novel enzyme system with CRISPR-like repeats, with only high-level direction from Anthropic scientists. Separately, the Life Sciences Verification Program, or LSVP, gives life science professionals access to Claude Mythos, Opus, and Sonnet models with a refined set of safeguards more permissive for biology-related work.

Current stage: independent wellbeing evaluation enters its review window

As of the Aug 25, 2026 announcement, the $5 million program has moved from concept to application stage. The next visible milestones are the September 21 deadline and the October 5 notification for full-proposal finalists. The program’s impact will depend on whether independent teams can produce open-source wellbeing evaluations that meet the stated criteria and whether other developers adopt them as shared measurement tools.