Some of what keeps AI safe is a good business. The rest is a public good no one pays for.

The people best at turning nothing into a working product are mostly pointed at ad tech and enterprise software. Meanwhile the labs building frontier AI expect systems that beat humans at most cognitive work within a few years.

Some of the safety this needs is a real business, and Juniper Ventures backs those founders. But the rest is neutral, shared infrastructure: independent testing, open biosecurity screening, the security and standards everyone relies on and no one owns. It has no business model, so almost no one is building it.

That is the gap Juniper Foundation funds. We pay proven founders to switch problems and go build it.

Others grow researchers. We fund builders to ship.

Excellent programs already train the research talent for AI safety. Almost none put proven founders — people who ship — onto the non-commercial infrastructure that has to exist and that the market will not pay for. That is the gap we fund.

We keep a running catalog of what is missing. Every brief says what to build, what it takes, where the money comes from, and the first 90 days. It is how we think, and how we hold the work to a standard.

A grant, and the help to use it.

We don’t re-test whether you can build. We select for fit.

The programs below already vetted your ability to ship, so we skip that. What we select for instead:

Once we know these, we move fast — often a decision within a week. But the bar is mission fit, not speed.

Outcomes, not headcount.

We judge the program by what gets built and used, not by how many fellows we fund. We track organizations launched onto a real safety, biosecurity, or security gap; tools, tests, and standards shipped and adopted by the people who need them; and the follow-on grant and mission funding the work attracts after us.

Every fellow publishes once a month, so the progress is public and the work stays honest.

Two asks.

1. Go full time. You leave your job or your startup. This is the work, not a side project.

2. Publish one post a month about what you are building and learning. It keeps the work honest and pulls more builders in.

That is it. Your equity, IP, and cap table stay yours.

Proven builders. No background in the field required.

If one of these programs accepted you — or you’ve raised venture funding — we skip re-vetting whether you can build.

You don’t need to know anything about AI safety or biosecurity — most fellows won’t. You need to be able to ship, and to actually care about the problem. We connect you with the advisor network for the rest. Not an alum? Get a referral from someone we trust, or apply anyway and tell us why. We read everything.

Exited founders, funding the field.

The fellowship is a program of Juniper Foundation, the nonprofit arm of Juniper. We are exited founders; we publish the AI Assurance Technology Report; and Juniper’s work is backed by people including Reid Hoffman, Wojciech Zaremba, Eric Ries, Geoff Ralston, and Jan Leike.

Grants come from the Foundation. Fellows keep their equity, IP, and cap table. This is a grant program, not dealflow for a fund.

Five steps.

The application.

Seven fields. We already trust whoever vetted you — a program or your investors. No idea yet? Say so and apply anyway.