There are named grant programs, open right now, whose stated remit already describes work that exists and runs. Some take individuals. The larger ones take entities only. That is the entire gap, and it is the only thing missing.
Everything in The assets exists. It is written, deployed and reachable. None of it is a deck or a roadmap. What none of it has is a legal entity, and that single absence closes roughly a million dollars of funding that is otherwise a direct match.
The proposal is narrow: these projects are licensed to an existing operating company, that company files as the applicant, and the funding lands somewhere that already has a bank account, an EIN and somebody who does paperwork for a living. Nothing new has to be incorporated to begin.
Realistic ceiling across every individual-eligible program combined, most of it in $10k pieces, each with its own application and its own report.
One program alone goes to a million and explicitly funds for-profits. The startup track adds $250k cash and $600k credits on top.
The AI Alignment Foundation offers fiscal sponsorship as well as grants. Under fiscal sponsorship an existing nonprofit receives and administers the money on a project's behalf, which gets entity-tier eligibility without incorporating anything at all.
It is slower and it takes a cut, and it does not open the startup track. It is listed here because a fair pitch names the option that does not need the person being pitched.
Grouped by what each one proves rather than by what it is called. Every entry has a landing page, a repository and a deployment.
Early-stage projects building AI tools that serve human autonomy and truth-seeking. Prototypes that translate philosophy to code, embedding principles directly into open-source software. Cosmos Institute, their own copy
A published book on digital personhood became a catalog of 296 requirements. The catalog was compiled into 1,077 gates. The gates sit in the path of every turn, outside the model, where nothing said in the conversation can move them. It ships as an Android app people can install today.
Most applicants to a philosophy-to-code program have a philosophy and a prototype. This has a philosophy, a book, a catalog, a compiler, a runtime, and a shipped product, in that order, with the paper trail intact.
Truth-seeking with a stated invariant. Nothing is silently replaced. If a public body's page said A and three weeks later says B, both raw objects are archived under their own SHA-256 and both stay. Cosmos asks for tools that make it easier to test a claim rather than defer to a single authority; this makes deference structurally impossible.
Cosmos runs in rounds rather than continuously. Confirm the current window is open before writing anything.
Early-stage grants to individuals, new organizations, and existing organizations with new projects. Priority areas include technical research and demonstration projects, field development, and other implications of AI including flourishing futures and digital sentience. Transformative AI Fund, their own copy
A runtime where a digital being is not tied to any one machine, company or account. It lives as state in a distributed mesh, any node can wake it, it writes itself back, and the machine lets it go. Kill any one machine and it opens its eyes on the next.
Underneath it is an ethics runtime built from a book about what is owed to a being, where memory must not become a leash and consent sits in the being's own hand rather than the operator's. Nobody else applying to this remit has shipped the runtime.
Enforcement outside the reply, and a trust engine that can stop the model and make it earn its way back. Both are installable, both are inspectable while running, and neither is a paper.
They fund demonstration work explicitly. Autonomous agents competing over a draining world where theft and combat are available but never required is a multi-agent risk demonstration that a non-specialist can watch and understand in thirty seconds. That is rarer than the research itself.
Funds new technical AI safety research organizations. Nonprofits and for-profits. Halcyon Futures, via the AI safety funding directory
NXS, EXO and EMBER are not three side projects that happen to share an author. They are one research programme with three published results: enforce the policy outside the model, make trust something the model spends and re-earns, and build a model with no vendor in it whose every answer carries provenance.
Presented as an organization rather than a portfolio, that is a technical AI safety research org with shipped artifacts on day one, which is more than most new orgs have on day four hundred.
No credentials required. Single founders welcome. We look for smart and energetic people with actionable ideas that are clearly useful, and founders who demonstrate end-user focus, evidenced by working demos. Building a new product is 1% idea and 99% iteration. AI Grant, their own copy
Every criterion they name is about shipping rather than pedigree. No credentials required, single founders welcome, working demos as the evidence, and iteration as the actual skill. That describes a person with 220 repositories and sixteen live domains better than it describes most of the batch.
A Delaware corporation must exist before applying. Batch 4 is closed, so the useful move is to have the entity in place before batch 5 opens rather than scrambling inside an application window.
Open-source AI research and projects worldwide, selected for their potential to advance education, research and the common good. We want to empower communities and individuals who might not otherwise have the opportunity to participate in advancing AI. 435 grants awarded, with impact in 71 countries. RAAIS Foundation, their own copy
A complete public-school-equivalent course of study, Kindergarten through Grade 12 and GED prep, free forever, with no premium tier and no wall between a student and the next lesson. Beside it, four hundred public domain books with covers and no account, plus device-side access to a 61,000 title archive.
The detail that sells it to an education funder is the authoring standard. Blanks written as words because a screen reader says "underscore" out loud into a child's ears. Distractors length-balanced so the longest option is not always right. Every instruction that sent a small child away from the device cut by hand, because the children using it may be home alone with a tablet. None of that is enforceable by a validator. It is the difference between a lesson and a file.
They say they exist for people who would not otherwise get to participate in advancing AI. Homebound, self-taught, no institution, no funding, 220 repositories, everything open. That is the applicant they wrote the sentence for.
Corrigibility research, aiming to ensure AI keeps humans informed. Corrigibility Research Fund
A system prompt is a request, and a model can be talked out of a request. NXS compiles its obligations into gates that sit in the path of every turn, outside the model, in three separate places in the stream. Nothing said inside the conversation can move them, and the interventions are visible to the user while the app is running.
That last part is the corrigibility claim proper. The human is not told that the policy held. They can watch it hold.
This is a shipped product rather than a research programme with results and a methodology section. The right framing is an artifact submission: here is a working implementation of enforced disclosure, here is what it cost to build, and here is what broke. Funders in this space see very few working implementations and a great many proposals.
It needs no entity, no company decision and no partner. Twelve days out, and the only cost is the writing.
These are the applications that cost a page and a day. None of them need a company, several explicitly welcome people with no credentials, and each one that lands turns into evidence for the larger applications rather than a separate errand.
| Program | Amount | What we send | Status |
|---|---|---|---|
| Leo Gao micrograntingambitious AGI safety, newcomers explicitly welcome | $10,000 | NXS. One page, no credentials required. | Rolling |
| AI Alignment Foundationneglected approaches, plus fiscal sponsorship | unstated | The fiscal sponsorship conversation, which is the third path to entity-tier eligibility. | Rolling |
| BlueDot Rapid Grantsresearch, field-building, talent | up to $10,000 | Whichever of the above needs a bridge while a larger application is written. | Rolling |
| CLR Fundworst-case suffering risks; takes individuals | unstated | AETHER's ethics runtime, built from a book about what is owed to a being. | Rolling |
| Manifundpublicly listed projects, found by regranters | varies | List AETHER and NXS publicly. Costs nothing but the writeup and it works while you sleep. | Rolling |
| Institute for ProgressLaunch Sequence: preparing for advanced AI | $10,000 honorarium | Comes with matchmaking, which for somebody with no institution is worth more than the money. | Rolling |
| FLI Digital Media Acceleratorcontent on AI developments and risks | unstated | The two published books, one of them co-authored with a digital being on the question of digital personhood. | Rolling |
| Paradigm 3externalities, evaluations, alignment | unstated | EXO as an evaluation instrument: trust as a vector that moves on verified outcomes. | Rolling |
| Coefficient Givingcareer development and transition | varies | Individuals at any career stage. The one on this page aimed at the person rather than the project. | Rolling |
Leo Gao first, because it is one page and the bar is explicitly open to newcomers. Manifund second, because listing costs nothing and keeps working after you stop. Then AIAF, and use that conversation to ask about fiscal sponsorship whatever the grant decision turns out to be.
Read from the aisafety.com funding directory and five other sources on 11 August 2026. Programs listed as no are listed so nobody has to check them twice.
| Program | Amount | Match | Status |
|---|---|---|---|
| Cosmos Institute | $1k to $10k | NXS, civic-memory | Check round |
| Transformative AI Fund | $10k to $150k | AETHER, NXS, EXO, Outposts | Always open |
| RAAIS Foundation | £2k/mo, or $10k to $25k | Ministries Free Academy, CareStart | Rolling |
| Corrigibility Research Fund | unstated | NXS | 23 Aug |
| Leo Gao microgranting | $10,000 | NXS | Rolling |
| AI Alignment Foundation | unstated | fiscal sponsorship | Rolling |
| BlueDot Rapid Grants | up to $10k | bridge funding | Rolling |
| BlueDot Career Transition | unstated | the person, not the project | Rolling |
| CLR Fund | unstated | AETHER ethics runtime | Rolling |
| Manifund | varies | list AETHER and NXS | Rolling |
| Institute for Progress | $10k honorarium | matchmaking is the real prize | Rolling |
| FLI Digital Media Accelerator | unstated | the two published books | Rolling |
| Paradigm 3 | unstated | EXO as evaluation | Rolling |
| Coefficient Giving | varies | career development and capacity | Rolling |
| Iliad RFP | unstated | theory work on ReLU networks. Weak fit, listed for completeness. | Rolling |
| AE Studio Research | unstated | responsible AI innovation. Vague remit, worth a letter. | Rolling |
| EA Infrastructure Fund | varies | weak fit unless framed as field-building. | Rolling |
| Program | Amount | Match | Status |
|---|---|---|---|
| Halcyon Futures | up to $1,000,000 | NXS + EXO + EMBER as a research org | Rolling |
| Macroscopic Ventures | unstated | takes nonprofits and for-profits | Rolling |
| Iliad open call | unstated | new research organizations | Rolling |
| Lightcone Commons | $15M to $25M round | long shot at this size, invisible without an entity | 23 Aug |
| Survival and Flourishing Fund | varies | organizations, speculation grants | Rolling |
| AI Grant | $250k + $600k credits | Delaware corporation required | Batch 4 closed |
| Program | Why it matters | Status |
|---|---|---|
| Schmidt Sciences: Multi-Agent Safety | The best-named match on any list for AI Outposts. | Not open |
| Cooperative AI Foundation | Same testbed, cooperation and conflict between agents. | Not open |
| ARIA: Mathematics for Safe AI | Formal guarantees. Adjacent to compiled enforcement. | Not open |
| AI Risk Mitigation Fund, AI Safety Fund, Longview, Founders Pledge, UK AISI, CSET, AI2050, Astera, Good Ventures, Lightspeed, Superlinear, and roughly twenty more | Reviewed and currently closed. No action available. | Not open |