2026 INTERPRETABILITY RFP - SCHMIDT SCIENCES

Cookie Preferences Center Cookies are small bits of data stored on the device (computer, mobile phone, tablet, or any other mobile device) that you use to access any of our websites (the Sites), so we can recognize repeat users. Each cookie expires after a certain period of time depending on what we use it for. In our Cookies Notice, we use the word “cookies” as a catchall term to include not only cookies, but also other technologies such as pixels, web beacons and page tags. Like most commercial websites we use both first and third party cookies on our Sites. We use cookies for several reasons. Below we list and describe the various types of cookies we use on our Sites. Depending on your geographic location, if you have signed up for services or filled out a sales contact form on our website we may share your email address in hashed form with marketing vendors to present personalized advertisements such as display, videos, and content on the web or connected devices. You can opt out via this portal. Our Privacy Notice Functional Cookies On These cookies collect information about your choices and preferences, and collect information about your use of the Sites and Services which enable us to improve functionality. For example, these cookies show us which are the most frequently visited pages on the Sites, allow us to present the Sites according to the settings you selected, help us record any difficulties you have with the Sites, and show us whether our advertising is effective or not. In some cases these cookies allow us to see the overall patterns of usage on the Sites, rather than the usage of a single person (except where that is part of the necessary function of the cookie). • Required Cookies On Certain cookies are necessary for this Site to securely operate, and for delivery of the Services you have requested. These cookies are limited to providing functionality, stability and security. You will not have the ability to reject these essential cookies in accordance with applicable data protection law. • Advertising Cookies On We may use cookies to market our services to you on third party websites or to measure and track the effectiveness of our marketing efforts so that we can tailor our advertising more effectively. We also contract with third party advertising networks that may collect IP addresses and other information from web beacons on our websites, emails, and on third party websites. Ad networks follow your online activities over time by collecting website navigational information through automated means, including through the use of cookies. They use this information to provide advertisements about products and services that may be of interest to you. You may see these advertisements on other websites. Confirm My ChoicesReject allAccept all EnglishDanskDeutschEspañolSuomiFrançaisItaliano日本語NederlandsNorskPortuguês - BRрусский языкSverigeTürkçe简体中文 en Powered by We value your privacy We and our third party partners may use cookies and similar technologies on this site to analyze usage, optimize our services, personalize content, tailor and measure ads and keep this site secure. Privacy Notice Cookies Notice Continue EnglishDanskDeutschEspañolSuomiFrançaisItaliano日本語NederlandsNorskPortuguês - BRрусский языкSverigeTürkçe简体中文 en Customize Skip to Main Content Skip to left navigation Schmidt Sciences 2026 Interpretability RFP Opens Mar 16 2026 12:00 PM (EDT) Deadline May 26 2026 11:59 PM (EDT) Description Request for Proposals: AI Interpretability Schmidt Sciences invites proposals for a pilot program in AI interpretability. We seek new methods for detecting and mitigating deceptive behaviors from AI models, such as when models knowingly give misleading or harmful advice to users. If this pilot uncovers signs of meaningful progress, it may unlock a significantly larger investment in this space. Core Question and Overview --- Can we develop interpretability methods that (1) detect deceptive behaviors exhibited by LLMs and (2) steer their reasoning to eliminate these behaviors? Successful tools will generalize to realistic use cases, moving beyond typical academic benchmarks and addressing concrete risks arising from deceptive behaviors. Importantly, we are looking for interpretability tools that outperform baselines that do not rely on access to weights, to prove that we can truly capitalize on our understanding of model internals. We define a scope of research in the [Research Agenda](https://docs.google.com/document/d/1fEaeRbx1A8DtzeV97LnHWLo5l