Other Efforts
Some of FLF’s work doesn’t take the form of starting new organizations. Instead, it includes research collaborations, individual scholarship, and short-term support for nascent projects that may grow into more — all aimed at furthering our priorities.
Example 1 — Community Notes research collaboration
FLF is supporting a university research collaboration focused on advancing Community Notes — the crowd-sourced fact-checking system first deployed on X and now being adopted or tested by Meta, YouTube, and others. Some example projects of this work include Reinforcement Learning from Collective Feedback (RLCF), AI-assisted tools for writing and rating notes, and more intelligent allocation of notes to raters. We see Community Notes as one of the most promising live examples of epistemic infrastructure, and we expect its impact to grow significantly as AI is more deliberately integrated into the loop.
Example 2 — Fiduciary AI research
FLF is supporting academic research towards standards and certification for fiduciary AI — work aimed at ensuring AI systems can be required to act in their users’ best interests, especially as those systems take on tasks beyond what users can meaningfully comprehend or supervise. The work includes basic research into viable approaches to standardization, expert convenings, and exploratory steps toward establishing a multi-stakeholder organization (modeled loosely on existing self-regulatory bodies in finance) that could host and enforce such standards. This complements other work we’re pursuing on fiduciary AI — including exploring a software layer that would monitor a user’s interactions with AI systems and chime in when those interactions don’t appear to be in the user’s best interests.
Example 3 — Nascent project support
FLF is also making small, short-duration grants — typically from one to three months — to nascent projects in our priority areas that may grow into full-fledged initiatives. These early awards give a team space to clarify their plan, test core assumptions, and decide whether to scale. They can also help us decide whether to back a larger effort down the line. One current example is a team working in AI for coordination that is figuring out their areas of focus and theory of change while quickly generating MVPs of technologies they might want to develop further, and testing them in their own work.
