Anthropic Launches $5 Million Grant Program to Evaluate AI's Impact on Wellbeing

Anthropic Launches $5 Million Grant Program to Evaluate AI's Impact on Wellbeing Anthropic has announced a $5 million grant program to fund independent, open-source evaluations of how AI models affect user wellbeing. Anthropic is launching a $5 million grant program offering direct funding, model access, and technical support for independent researchers. Grantees will create open-source evaluations and benchmarks to help the AI industry measure its impact on user wellbeing. Wellbeing is particularly difficult to evaluate because assessing risks like self-harm or disordered eating requires multi-turn conversational context rather than evaluating single responses in isolation. The program seeks participation from clinicians, psychologists, and methodologists, with applications due September 21 and notifications for full proposals by October 5. Grant Program Overview Anthropic is dedicating $5 million in direct funding, technical support, and model access to independent researchers building open-source evaluations. Grantees will work independently to publish benchmarks and evaluation tools that any developer across the AI industry can use to gauge model effects on users. Evaluating Contextual Wellbeing Challenges Unlike standard model behaviors that can be assessed from a single prompt and reply, evaluating wellbeing requires extensive conversational context. For instance, signs of distress or self-harm may only surface across extended interactions, and responses such as diet advice can become harmful if a user has a history of disordered eating. Interdisciplinary Goals and Deadlines Anthropic is encouraging collaboration across disciplines, inviting clinicians, psychologists, and methodologists to help formulate industry standards. Alongside the grant, Anthropic's Safeguards team has published guidance on building rigorous evaluations. Applications for the program close on September 21, with selected candidates notified by October 5 to submit full proposals.

Anthropic设立500万美元资助计划以研究AI对用户福祉的影响 Anthropic宣布启动一项总计500万美元的资助计划，支持独立研究人员构建开源评估工具，以衡量人工智能系统对使用者心理福祉的影响。 该计划提供直接资助、模型使用权与技术支持，要求成果完全作为开源项目发布供全行业使用。 评估福祉比普通任务更复杂，需要依赖长对话的上下文判断潜在风险，例如应对自残倾向或进食障碍历史。 Anthropic希望吸引临床医生、心理学家和方法学家参与，申请截止日期为9月21日，入选全案阶段的通知时间为10月5日。 资助计划与开放目标 Anthropic投入500万美元资助独立研究，旨在帮助AI行业衡量模型如何影响用户福祉。受资助者将保持完全独立的工作方式，除资金外还将获得模型使用权限和技术支持，所有产出需作为开源项目公开发布，以便任何开发者采纳。 评估用户福祉的特殊挑战 AI正成为许多人的对话伙伴与情感支持来源，但行业目前缺乏明确的行为规范，例如面对陪伴需求或心理健康危机时模型应如何应对。与仅凭单次回答即可判定准确性的普通任务不同，福祉评估极其依赖上下文。例如用户在危机初期可能不会立即透露自残念头，普通的节食建议也可能对有进食障碍史的用户造成潜在伤害。 专业参与与申请流程 为应对这一动态演变的挑战，Anthropic期望临床医生、心理学家与方法学专家加入该领域。其安全团队已发布关于构建高标准评估体系的指导意见。该计划申请截止日期为9月21日，获选提交完整提案的申请者将于10月5日前收到通知。

We’re launching a $5 million grant program to fund independent research into how AI impacts users’ wellbeing. The program will provide direct funding, access to our models, and technical support to grantees building open-source evaluations that help the AI industry measure how our models affect those who use them. Grantees will work fully independently, and will publish their work as open-source projects that any developer can make use of. AI systems have become central to how many people work, learn, and solve problems. They’ve also become conversational partners and can be [sources of emotional support](https://www.anthropic.com/news/how-people-use-claude-for-support-advice-and-companionship) during difficult times. But as an industry, we are still working towards developing clear standards for how models should behave in these conversations, for example, when a user begins to seek companionship from a model, or uses AI to navigate a mental health crisis. Furthermore, wellbeing is a particularly difficult area to evaluate. For most model behaviors, we can look at a single answer and determine whether it is accurate and appropriate. But assessing wellbeing requires much more context. For example, 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. And a response that might be reasonable in one context might be harmful in another. For example, Claude might give advice on balanced diets and workout routines to a user who asks about losing weight, but if the user has demonstrated a history of disordered eating, that response could be inappropriate, and potentially actively harmful. We work to [develop safeguards](https://www.anthropic.com/news/protecting-well-being-of-users) to identify such conversations and help ensure Claude responds appropriately, and we [publish research](https://www.anthropic.com/research/claude-personal-guidance) into the types of conversations people have with Claude to better inform how we develop our safeguards, how we evaluate them, and other measures we can take to protect users’ wellbeing. But these are nuanced considerations, and the stakes are significant. The right approach will need to evolve alongside our models and their uses. By funding the creation of independent evaluations and benchmarks of user wellbeing, we hope to invite more people to lend their expertise to this emerging and critical field, including clinicians, psychologists, methodologists, and others. As part of this program, we’re [sharing guidance](https://www-cdn.anthropic.com/files/4zrzovbb/website/5ecb637cb206057cb93cf4a9e72e843fda5e9892.pdf) from our Safeguards team on what we believe makes a wellbeing evaluation rigorous enough to build on, along with the common challenges that can limit an evaluation’s usefulness. In brief, we’re seeking evaluations that: To learn more about the grant program and apply, [see our application form](https://docs.google.com/forms/d/e/1FAIpQLSfmUDtpfg-ztmQPkCbwx-VAe_urX48sthbqs8GmZUT9irRSDQ/viewform?usp=dialog). For more on building strong wellbeing evaluations and benchmarks, [read our guidance](https://www-cdn.anthropic.com/files/4zrzovbb/website/5ecb637cb206057cb93cf4a9e72e843fda5e9892.pdf). Applications are due by September 21; applicants who are selected to submit full proposals will be notified by October 5.