Hide table of contents

Below is a fictionalized list of grants the Long-term Future Fund (LTFF) narrowly rejected or accepted in the last 6 months. We aim to broadly convey which grants we barely rejected, while anonymizing any individual one. It follows our previous two posts on marginal grantmaking, from August and November of 2023. We hope this is useful to potential donors and other community members.

As the grants we’ve fictionalized are grants LTFF narrowly accepted or rejected, additional funding to LTFF will allow us to roughly pay for projects of similar or slightly lower quality and approximate cost-effectiveness. If that appeals to you, please consider donating here (alt : every.org).

Grants

6-month stipend to continue AI safety independent research through temporal understanding and deception detection ($60,000)

This 6-month project investigates situational awareness in language models through temporal understanding, aiming to develop realistic scenarios for deceptive alignment by exploring future-past representations, training distribution-shift-based backdoors, and evaluating targeted safety interventions. The investigator has published research on multimodal robustness benchmarking, automated debiasing, and temporal convolutional networks in major ML venues, while demonstrating organizational leadership by scaling a technical volunteering program from 20 to 150+ practitioners serving 35 partner organizations.

10-month stipend for advancing biosecurity through DNA synthesis screening standards ($105,000)

This 10-month project strengthens biosecurity safeguards by developing technical security standards for emerging DNA synthesis platforms, creating nucleic acid screening policies, and building consensus among industry stakeholders through national and international standards bodies. The investigator has published multiple papers on biosecurity governance and DNA screening in venues like AAAI[1] and Global Biosecurity, directly advised major DNA synthesis companies and government agencies on security implementations, and achieved concrete policy outcomes including proposed screening guidelines adopted by international consortiums and security recommendations implemented by commercial manufacturers.

Stipend for international law expert to finish a book advancing AI governance frameworks ($25,000)

This 4-month project develops foundational frameworks for international AI governance through writing a book on catastrophic risk law, contributing handbook chapters on AI treaties, and drafting model governance structures, while directly advising key international bodies on AI policy. The investigator has authored influential policy papers on institutional models for AI governance, contributed to major international principles on intergenerational equity, shaped executive-level AI policy directives, and built academic-policy networks across multiple international organizations, while developing the first legal framework for governing catastrophic AI risks through Yale University Press. The grant will be a stipend to pay them to finish the book; earlier sections were funded elsewhere.

Stipend for empirical analysis of coalition-building dynamics within AI governance ($40,000)

This 8-month project examines effective AI governance through data-driven analysis of industry influence, regulatory dynamics, and alliance-building strategies, combining systematic collection of corporate metrics with game-theoretic modeling of institutional cooperation. The investigator has developed a long-form interview series exploring emerging technology risks with 50+ episodes, led technical research teams studying multi-agent AI systems, and scaled student organizations focused on AI safety from 20 to 200+ members, while producing policy research on institutional decision-making. The project will be mentored by respected AI policy researchers in an academic setting.

Tuition and partial stipend subsidy for 1 year of doctoral research on security of advanced biotechnologies ($40,000)

This thesis project analyzes emerging biotechnology vulnerabilities through systematic risk assessment of synthesis technologies, developing a timeline-based framework for evaluating bioterrorism pathways, and creating evidence-based policy recommendations for counterterrorism interventions. The investigator has demonstrated strategic leadership in national biosecurity policy implementation, developed computational tools for security applications, and built cross-sector health security networks while maintaining research positions at leading institutions. (The project is partially funded elsewhere)

9 months’ stipend for developing a benchmark for goal-directedness in large language models ($65,000)

We cover 3 people’s stipends for three months to develop a benchmark for assessing goal-directed behavior in language models through simulated environments, adversarial testing, and distributional shift scenarios to measure model robustness and capability trends. The investigators have developed evaluation frameworks showing capability differences across architectures, contributed 25,000+ test cases to public benchmarks, published in peer-reviewed venues, and expanded their research initiative from 4 to 7 members while delivering technical training.

Part-time project for replacing neural networks with interpretable programs ($6,000)

This part-time project aims to create transparent, programmatic replacements for sparse autoencoder neurons in language models by developing symbolic representations in Python, evaluating their predictive accuracy, and measuring their impact on model performance through cross-entropy loss. The investigator has published research on neural network optimization and language model architectures in major ML venues, developed open-source interpretability tools, and demonstrated expertise in language model training and low-level optimization methods across industry and research settings. The investigator has a full-time job in industry and our stipend only covers weekends working on this research project.

One year of costs for an AI safety talent and community-building project in Southeast Asia ($50,000)

This project aims to strengthen the AI safety field across Southeast Asia[1] by expanding a proven community-building model through targeted social media outreach, local events, and educational programming for high-potential individuals. The organization has grown active membership from 20 to 250+ participants, delivered 10 technical workshops with 85% satisfaction, and partnered with 10 institutions while achieving 45% conversion from initial contact to sustained engagement.

3 years of PhD work on an AI predictability framework to enable safer deployment ($175,000)

This project develops a framework for evaluating and monitoring AI system predictability through behavioral indicator assessment, human-AI interaction analysis across four manifolds, and scalable oversight mechanisms. The investigator has published research on AI system reliability and evaluation in major journals, conducted extensive safety testing of large language models, and contributed evaluation frameworks for assessing AI capabilities and risks while working with leading research institutions and safety organizations.

___

The above fictionalized examples illustrate grants that are at or just below our current funding threshold. If these projects seem worth funding to you, please consider donating to us here (alt: every.org).

  1. ^

    All proper nouns should be assumed to be fictionalized.

Comments3


Sorted by Click to highlight new comments since:

Thanks for sharing this, I have been really enjoying reading the comms coming from EAIF/LTFF lately! Here are some observations/feedback on the post:

Below is a fictionalized list of grants the Long-term Future Fund (LTFF) narrowly rejected or accepted in the last 6 months. We aim to broadly convey which grants we barely rejected, while anonymizing any individual one.
[...]
As the grants we’ve fictionalized are grants LTFF narrowly accepted or rejected

After reading the post, I feel like I really miss knowing which of these did or did not pass your bar. I understand that these are fictional examples that are all very close to the bar, but I think others might feel the same way. If you think it might be worth spelling out why you don't think it makes sense to give explicit verdicts.

The above fictionalized examples illustrate grants that are at or just below our current funding threshold

Or perhaps this means all of them were rejected? (It might just be my English, but it's not clear to me if something is at your funding bar, then you would accept or reject it)

On a related note, if all of these grants were rejected, would the applicants asking for 10-25% less funding would make them pass your bar? Do you often end up funding the "MVP" version of a project as opposed to the "mainline budget" they propose?

Executive summary: The Long-Term Future Fund shares examples of grants they narrowly accepted or rejected, illustrating their funding threshold and demonstrating that additional donations would support similar projects focused on AI safety, biosecurity, and related existential risk research.

Key points:

  1. Grant amounts range from $6,000 to $175,000, covering research stipends, PhD work, and community building projects.
  2. Focus areas include AI safety (interpretability, benchmarking, governance), biosecurity (DNA synthesis screening), and international policy frameworks.
  3. Most projects demonstrate strong prior expertise, institutional connections, and concrete track records of impact.
  4. Projects typically combine technical research with practical applications or policy implications.
  5. The fund seeks additional donations to support more projects at this threshold of quality and cost-effectiveness.

 

 

This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.

Curated and popular this week
 ·  · 7m read
 · 
This is a linkpost for a paper I wrote recently, “Endogenous Growth and Excess Variety”, along with a summary. Two schools in growth theory Roughly speaking: In Romer’s (1990) growth model, output per person is interpreted as an economy’s level of “technology”, and the economic growth rate—the growth rate of “real GDP” per person—is proportional to the amount of R&D being done. As Jones (1995) pointed out, populations have grown greatly over the last century, and the proportion of people doing research (and the proportion of GDP spent on research) has grown even more quickly, yet the economic growth rate has not risen. Growth theorists have mainly taken two approaches to reconciling [research] population growth with constant economic growth. “Semi-endogenous” growth models (introduced by Jones (1995)) posit that, as the technological frontier advances, further advances get more difficult. Growth in the number of researchers, and ultimately (if research is not automated) population growth, is therefore necessary to sustain economic growth. “Second-wave endogenous” (I’ll write “SWE”) growth models posit instead that technology grows exponentially with a constant or with a growing population. The idea is that process efficiency—the quantity of a given good producible with given labor and/or capital inputs—grows exponentially with constant research effort, as in a first-wave endogenous model; but when population grows, we develop more goods, leaving research effort per good fixed. (We do this, in the model, because each innovator needs a monopoly on his or her invention in order to compensate for the costs of developing it.) Improvements in process efficiency are called “vertical innovations” and increases in good variety are called “horizontal innovations”. Variety is desirable, so the one-off increase in variety produced by an increase to the population size increases real GDP, but it does not increase the growth rate. Likewise exponential population growth raise
 ·  · 25m read
 · 
Epistemic status: This post — the result of a loosely timeboxed ~2-day sprint[1] — is more like “research notes with rough takes” than “report with solid answers.” You should interpret the things we say as best guesses, and not give them much more weight than that. Summary There’s been some discussion of what “transformative AI may arrive soon” might mean for animal advocates. After a very shallow review, we’ve tentatively concluded that radical changes to the animal welfare (AW) field are not yet warranted. In particular: * Some ideas in this space seem fairly promising, but in the “maybe a researcher should look into this” stage, rather than “shovel-ready” * We’re skeptical of the case for most speculative “TAI<>AW” projects * We think the most common version of this argument underrates how radically weird post-“transformative”-AI worlds would be, and how much this harms our ability to predict the longer-run effects of interventions available to us today. Without specific reasons to believe that an intervention is especially robust,[2] we think it’s best to discount its expected value to ~zero. Here’s a brief overview of our (tentative!) actionable takes on this question[3]: ✅ Some things we recommend❌ Some things we don’t recommend * Dedicating some amount of (ongoing) attention to the possibility of “AW lock ins”[4]  * Pursuing other exploratory research on what transformative AI might mean for animals & how to help (we’re unconvinced by most existing proposals, but many of these ideas have received <1 month of research effort from everyone in the space combined — it would be unsurprising if even just a few months of effort turned up better ideas) * Investing in highly “flexible” capacity for advancing animal interests in AI-transformed worlds * Trying to use AI for near-term animal welfare work, and fundraising from donors who have invested in AI * Heavily discounting “normal” interventions that take 10+ years to help animals * “Rowing” on na
 ·  · 14m read
 · 
As we mark one year since the launch of Mieux Donner, we wanted to share some reflections on our journey and our ongoing efforts to promote effective giving in France. Mieux Donner was founded through the Effective Incubation Programme by Ambitious Impact and Giving What We Can. TLDR  * Prioritisation is important. And when the path forward is unclear, trying a lot of different potential priorities with high productivity leads to better results than analysis paralysis. * Ask yourself what the purpose of your organisation is. If you are a mainly marketing/communication org, hire people from this sector (not engineers) and don’t be afraid to hire outside of EA. * Effective altruism ideas are less controversial than we imagined and affiliation has created no (or very little) push back * Hiring early has helped us move fast and is a good idea when you have a clear process and a lot of quality applicants Summary of our progress and activities in year 1 In January 2025, we set a new strategy with time allocation for our different activities. We set one clear goal - 1M€ in donations in 2025. To achieve this goal we decided: Our primary focus for 2025 is to grow our audience. We will experiment with a variety of projects to determine the most effective ways to grow our audience. Our core activities in 2025 will focus on high-impact fundraising and outreach efforts. The strategies where we plan to spend the most time are : * SEO content (most important) * UX Optimization of the website * Social Media ; Peer to Peer fundraising ; Leveraging our existing network The graphic below shows how we plan to spend our marketing time: We are also following partnership opportunities and advising a few high net worth individuals who reached out to us and who will donate by the end of the year. Results: one year of Mieux Donner On our initial funding proposal in June 2024, we wrote down where we wanted to be in one year. Let’s see how we fared: Meta Goals * Spendi