Contributors
0 of 1,000,000
Observations
0 of 1,000,000,000
Contribute to the Research
The research exists. The infrastructure is designed. What's missing is you.
Dee Williams on why this research matters.
Audacion AI Labs is building the post-deployment safety field at the moment it has to exist. The research is underway. The framework is built. The contributor model is live. The infrastructure to study what AI actually does after it leaves training is in motion, and it is being built in the open. We need partners and funders who understand that safe AI is not going to build itself.
What Your Funding Enables
The architecture is built. The research is underway. What your funding does is sustain the engine, expand the reach, and turn the science into something the world can use.
Sustain the Research
Keep the science engine running. Funding sustains the researchers, the analysis pipeline, the publication cycle, and the ongoing expansion of the behavioral drift taxonomy. This is the line item that keeps the post-deployment field alive.
Expand the Tools
Build the platform ecosystem that turns contributors into a research instrument: web portal, browser extension, data pipeline, and new surfaces for collecting observations from the people who use AI every day.
Spread the Word
Awareness is a research multiplier. Funding supports the billboards, podcasts, press relationships, and community work that bring new contributors and new findings into the lab. The more eyes on AI behavior, the better the science.
Fund the Operations
Administration, legal, compliance, and infrastructure. The unglamorous work that lets a research lab actually function. Audacion holds operations at or below 15% of total spend so that every dollar of funding is doing visible work.
Operations at or below 15% of total spend.
Power the Partnerships
Funding fuels the governance cycle: research feeds partnerships, partnerships feed training, training drives safer usage, safer usage produces better data, and better data sharpens the science. Every funded dollar compounds across the loop.
Why This Lab
We don't study AI in isolation.
We study it where it actually operates, inside real deployments, with real users, under real conditions. The published literature is two years behind the deployment reality. We are not.
We don't keep the research behind closed doors.
Our findings, taxonomy, and methodology are built to be public. AI safety cannot be a private discipline if it is going to matter at the speed AI is moving.
We don't treat citizens as the audience.
The teachers, nurses, parents, and operators who use AI every day are part of the research, not its audience. Citizen science is structural to how the lab works, not a marketing line.
We don't pretend pre-deployment evaluation is enough.
Roughly 98% of published AI safety research ends at deployment. We start there. The post-deployment surface area is where the next decade of safety findings will come from.
We don't take partner money in exchange for partner protection.
Editorial independence is non-negotiable. No partner reviews, edits, approves, or delays a research finding. This is the firewall that makes the science worth funding in the first place.
The AI Advisory Council
No other AI safety lab has structured a seven-frontier-model advisory council into its research process. The Council bridges operational practitioners, academic researchers, and the frontier models themselves, a research architecture no other organization in the field has built.
Ways to Support
Grant Funding
Audacion is actively pursuing and welcomes federal and foundation grant support, including NSF SBIR, Cartier Women's Initiative, CalOSBA, and aligned foundations. Directed grants toward specific PRISM pillars are encouraged.
Institutional Partnership
Multi-year institutional support from enterprises, foundations, and research institutions. Partners receive ongoing research access, advisory engagement, and the benefits described on the Partners page.
Our editorial independence is structural.
Institutional partners receive research access, advisory engagement, and recognition. They do not receive influence over what the lab studies, what it publishes, or how it reports findings. That boundary is not a policy. It is architecture.
Sponsorship
Named sponsorship of fellowships, the annual PEAQ Summit, recurring research convenings, and the lab's publication series. Sponsorship is recognized publicly while remaining outside the research review process.
Data Partnership
A different kind of support: contributing model access, compute, anonymized datasets, or operational telemetry that strengthens the research corpus. Data partnership is distinct from financial support and is governed separately under the independence policy.
Advisory
Senior expertise contributed in safety, research methodology, grants, policy, community building, or citizen science. Advisory engagement does not require financial commitment, but it does require time and direct involvement with the research agenda.
The Numbers
1,000,000
contributors
1,000,000,000
observations
10
year research program
5
research dimensions
58
in the taxonomy (20 original)
7
models on the AI Advisory Council
Your funding makes this real.
Start a conversation.
Prefer to call? Reach us directly at (424) 999-0548 or email [email protected].