Loading organizations…
Loading organizations…
Transluce is an independent nonprofit AI research lab headquartered in San Francisco, California. The organization was co-founded on October 23, 2024 by Jacob Steinhardt, an assistant professor at UC Berkeley's Department of Statistics, and Sarah Schwettmann, a research scientist at MIT CSAIL. The name "Transluce" means to shine light through something to reveal its structure, reflecting the lab's core commitment to transparency.
The lab's mission is to build open, scalable technology for understanding AI systems and steering them in the public interest. Transluce believes that as AI systems grow more capable, human oversight must scale alongside them · something that requires automated tools rather than purely manual inspection. Its initial technical releases included automated feature descriptions for neuron activation patterns, an observability interface for interrogating AI system internals, and a behavior elicitation agent capable of searching for specific behaviors in frontier models such as Llama-405B and GPT-4o…
Transluce believes that scalable democratic oversight of AI requires automated tools that can match the pace and complexity of modern AI development. Their causal chain is: (1) build open-source AI-driven tools that can automatically analyze and explain the internals and behaviors of large AI models; (2) put these tools in the hands of independent evaluators, governments, and civil society so that safety assessments are no longer controlled solely by commercial labs; (3) establish shared industry standards · through bodies like the AI Evaluator Forum · that normalize independent auditing; (4) use public audits and transparency to create accountability pressure that nudges AI developers toward safer deployment practices. By operating as a nonprofit that openly publishes its methods, Transluce aims to become a trusted, independent reference point that can credibly identify risks such as deception, hallucination, and misuse before they cause harm.
Manifund funding goal
Closed Jan 22, 2026
$500 minimum · $2M ideal
Fundraising data from grantmaking.ai, updated .
View on grantmaking.ai# Project summary
Transluce is a non\-profit AI lab working to ensure that AI oversight scales with AI capabilities\. This means developing novel automated oversight tools and putting them in the hands of AI evaluators, companies, governments, and civil society\. Reliable, independent evaluations of AI systems are key to both safety and innovation, and Transluce is building tools to make this practical at scale and across domains\.
So far, we have built scalable systems for monitoring AI agents, testing their behaviors, and interpreting their inner workings, and used these to study issues like sycophancy, self\-harm, and reward hacking\. Our systems have been used to build and improve popular agent evaluations like HAL and SWE\-bench, by developers to align frontier models like Claude 4, and by governments to evaluate risks to public safety\. We…
The project’s theory of change is that reliable democratic oversight of frontier AI requires scalable, AI-backed tools for understanding model behaviors, rather than ad hoc, closed-door testing by deploying labs alone. By using philanthropic funding to develop automated systems for monitoring agents, surfacing rare and harmful behaviors, and interpreting internal representations · and then making these tools available to independent evaluators, companies, governments, and civil society · Transluce aims to make rigorous third-party evaluations practical at scale. Publicly validated tools and widely adopted evaluations, in turn, are intended to create transparency and accountability pressures that push AI developers toward safer deployment practices.