Foresight Institute AI for Science & Safety Nodes 2026: Up to $100,000 Plus Compute and Office Space

Captured cycle: This page is retained for historical guidance. Confirm whether the program has reopened before planning an application.

Foresight Institute AI for Science & Safety Nodes 2026: Up to $100,000 Plus Compute and Office Space

Current status

The Foresight Institute official program page currently lists July 31 as the next application deadline for its AI for Science & Safety Nodes program. The page links to an Airtable application form and describes the Nodes as active physical hubs in San Francisco and Berlin. This page therefore describes the current 2026 opportunity, not a closed historical round. The deadline field above uses the single date published on the official page; applicants should check the form itself before submitting because the page does not state a submission time zone.

The program is broader than a conventional research grant. Foresight presents three connected forms of support: grant funding, office and community space, and local private compute. An applicant may ask for all three or select only the forms of help that match the project. The strongest fit is an AI-first project whose team wants to work as part of a Node, attend relevant gatherings, and collaborate with other researchers and builders. A proposal that only seeks cash, with no credible reason to use the spaces or community, faces a weaker fit because funding-only applications are considered only in exceptional cases.

Official program page: https://foresight.org/grants/grants-ai-for-science-safety/

Application form: https://airtable.com/appyVXc5SMPAvIKpP/pagzBRWeiG3HjH6Qn/form

At a glance

What the Nodes provide

Foresight describes the Nodes as physical hubs for AI-powered science and safety. The first support option is grant funding. Typical awards range from $10,000 to $100,000, and the organisation says it awards around $3 million in total each year. The published range is not a promise that every applicant can request or receive the maximum. The page says higher amounts are awarded in the AI safety-oriented areas, while longevity biotechnology and molecular nanotechnology projects generally receive smaller amounts.

The second support option is office and community space. Foresight strongly prefers applicants who plan to be active members of a San Francisco or Berlin Node. The reason is operational rather than cosmetic: the organisation says the value of the Nodes comes from peer collaboration and the work that happens between scheduled programming. A useful application should therefore explain which people will participate, where they can work, and what interaction with neighbouring projects would add to the plan.

The third option is local, private compute. Eligible projects may receive a compute budget or access to local compute resources, with particular interest in privacy-oriented work. The official page asks applicants to state how much compute they need and what they will use it for. A credible request should connect compute to a defined workload, such as training, evaluation, simulation, secure inference, or a reproducible scientific pipeline. A general statement that more compute would be helpful does not show why Node resources are necessary.

Applicants may request one, two, or all three kinds of support. The form should be used to make that choice explicit. If a project needs funding but not a desk, the proposal needs to make a convincing case for why a funding-only arrangement is an exception to the normal Node model.

Who can apply and what makes a project fit

The official FAQ says that individuals, teams, and organisations can apply. It welcomes both non-profit and for-profit organisations. A for-profit applicant must be prepared to explain why it needs grant funding. That explanation should address the public-interest value of the work and why ordinary commercial financing is not an adequate substitute; the page does not say that every for-profit project is eligible automatically.

The central subject-matter requirement is an AI-first approach. Foresight says it gives priority to projects that use AI as the primary engine for progress across its focus areas. This is a more demanding standard than adding a model to an otherwise unchanged project. Your proposal should state what AI makes possible, what result will be delivered, and how progress can be measured within the proposed period.

The page lists seven areas of interest:

1. AI for Security: self-improving defence systems, automated vulnerability discovery, formal proofs, red-teaming, and stronger digital infrastructure.
2. Private AI: confidential compute, privacy mechanisms, distributed trust, and local private compute stacks.
3. Decentralised and Cooperative AI: systems that cooperate, negotiate, align, and support peaceful human–AI coexistence.
4. AI for Science and Epistemics: tools, platforms, data infrastructure, forecasting, and systems that improve scientific progress or general sense-making.
5. AI for Neuro, Brain-Computer Interfaces, and Whole Brain Emulation: work using models to map, simulate, or understand biological intelligence.
6. AI for Longevity Biotechnology: AI applied to biostasis, replacement, gene therapy, exosomes, and related longevity research.
7. AI for Molecular Nanotechnology: AI for the design, simulation, construction, or assembly of molecular machines.

The evaluation criteria differ slightly by focus area. For AI safety projects, the page highlights impact on reducing existential risks from AI, feasibility within short AGI timelines, alignment with the listed areas, capability to execute, high-risk/high-reward potential, and a preference for open source. The page explicitly says that the AI safety criteria do not apply to the longevity biotechnology and molecular nanotechnology areas. An applicant in one of those areas should still explain the scientific case, feasibility, team capability, and fit with the relevant focus area rather than borrowing an AI-safety argument that does not match the work.

How to apply for the current cycle

Start at the Foresight program page and open the linked Airtable form. The application route is not a separate grants portal: the official page directs applicants to that form. Before submitting, decide whether the request is for funding, office space, compute, or a combination. Describe the Node location that is practical for the team and be direct about whether active in-person participation is realistic.

The proposal should then make five things easy to find:

• the problem and the specific result the project will produce;
• the role AI plays as the primary method or engine of progress;
• the focus area or areas that best match the work;
• milestones and deliverables that can be completed in the next one to three years;
• the budget, including the requested grant amount and any compute or space needs.

For compute, state the amount needed and the purpose, as the official page requests. For a multi-year project, show how work will be divided into milestones. For a for-profit application, include the reason grant support is necessary. For an open-source project, explain what will be released and when; Foresight states a preference for open-source projects unless there is a specific reason not to share the work.

The official funding terms also affect preparation. Grants are usually paid in one lump sum. Larger projects spanning multiple years may instead receive payments in tranches, with later payments dependent on successful completion and reporting of earlier milestones. Foresight may cover overhead up to 10% of direct research costs when those costs directly support the funded work. Successful applicants must pass due diligence, which can include confirming connections to Foresight Institute, disclosing ongoing criminal proceedings or bankruptcy, and sharing an itemised budget, project plan, and organisational documents.

Submit through the form by the listed July 31 deadline. The official page estimates about two months of review after each application deadline, although smaller funding requests may move faster. Applicants may request fast processing, but Foresight says it may not be able to honour that request. Proposals are first reviewed in-house for fit and quality. Strong submissions may then go to technical advisers, followed by written questions or a short call if the proposal advances. Foresight says it cannot provide individual feedback to unsuccessful applicants, so keep a copy of the submission and maintain a clear record of what could be improved for a later opportunity.

A practical preparation checklist

Before opening the form, write a short project summary that names the problem, the AI-enabled approach, and the first measurable result. Map the work to one or two focus areas instead of listing every area. Set out a milestone sequence that a reviewer can test: what will be built, evaluated, or demonstrated, what evidence will count as progress, and what the team will do if an approach fails.

Prepare a budget that separates direct project costs from any requested overhead. If compute matters, include workload, duration, privacy requirements, and an explanation of why local private resources are useful. If Node participation matters, identify the people who will be present in San Francisco or Berlin and describe the collaborations or events that would improve the work. If participation is not feasible, explain why the project qualifies for the exceptional funding-only route rather than implying that the location requirement does not exist.

Finally, confirm the exact fields and attachment requirements shown in the live Airtable form. The public page gives the program rules but does not publish every field in a static checklist. Do not assume that a previous submission, an old deadline, or a generic grant template matches the current form. The official page currently points to July 31 as the next deadline, so use that date for planning and recheck the official page and form immediately before submission.

Bottom line

This is a current Foresight Institute Node opportunity with a July 31 deadline, not a permanently open monthly grant listing. It can support an AI-first science or safety project with money, private compute, office space, and a working community in San Francisco or Berlin. The most persuasive application will show a clear reason for each requested resource, concrete one-to-three-year deliverables, a close match to a listed focus area, and a credible plan for active Node participation. The program accepts individuals, teams, non-profits, and for-profits, but the last group must justify the need for grant funding. Use the official page and its linked Airtable form for the final submission instructions.