Our 40th Year Impact Report is live. Plus: Vision Weekend USA details, and Secure & Sovereign AI talks.
In this newsletter:
Next Week: A sold-out Vision Weekend returns to San Francisco
Our 40th Year Impact Report: What our community said, and what the numbers show.
New Grantees across AI safety and AI-first science
Seeking Grant Advisors: Request for proposals in AI for science & safety
Leadership transition: Allison Duettmann joining our Board of Directors
Plus: Videos from our Secure & Sovereign AI Workshop are live
40th Year Impact Report: Now Live
What makes Foresight unique? In our 40th year, we took the opportunity to ask our community about the impact of Foresight. Over 300 people answered, including 12 from the founding cohort of 1986–1990. Their collective responses, along with data from across our programs, make up our 40th Year Impact Report.
This report is our reflection on forty years of Foresight, and where we stand in 2026.
It covers each of our programs — grants, fellowships, workshops, seminars, prizes, Vision Weekends, Existential Hope — with this year’s impact numbers and the outcomes our community reported.
Thank you to everyone who lent their voices to this report. This is your story as much as ours.
P.S. We’ll have printed copies at Vision Weekend in San Francisco next week. Come find one.
Vision Weekend USA: Only a Week Away
A sold-out Vision Weekend returns next week to San Francisco!
Expect our main 40th anniversary celebration in the US, expert talks across seven tracks, a $10K pitch competition, sci-fi book swap, behind-the-scenes lab tours, and more. See the full program.
Concert pianist and computer scientist Serene is set to perform Friday night aboard the USS Hornet aircraft carrier. Serene is a Bösendorfer Artist, a former Google engineer, and the creator of Snowflake, a censorship-circumvention protocol.
All Access ticket holders are invited to join our 40th anniversary celebration, including lab tours at E11 Bio, Longshot, CalWave, Dolphin Labs, Science Corporation, plus a Cypherpunk party at our AI Node.
Ticket holders: Don’t miss our Founder & Funder Mixer and two competitions
On Sunday, Nov 15, we’re hosting a Founder & Funder Mixer at Lighthaven: a curated evening for founders, researchers and investors working on secure AI, nanotech, neurotech, space and energy, and longevity biotech.
Vision Weekend ticket holders have the mixer included, however, you must register on Luma to be matched with funders and founders! Non-ticket holders can apply to attend here.
The mixer also features two competitions:
Vision Weekend Idea Prize Competition 2026: open only to Vision Weekend participants, and the winner receives the $10,000 Vision Weekend Idea Prize. Enter your idea.
Founder Pitch Competition: open to founders of startups, frontier research organizations (FROs) and public benefit corporations (PBCs) who are raising funds. Selected founders get to pitch in front of funders. Apply to pitch.
Vision Weekend is sold out. Thank you for the incredible interest! All tickets have been claimed, but spots sometimes open up last minute. Join the waitlist and we’ll contact you if a ticket becomes available.
Vision Weekend USA is powered by: Protocol Labs, EigenLabs, Paradromics, Retro Biosciences, FUTO, d_model, and Convergent Research.
Secure & Sovereign AI Workshop in Berlin: Talks Now Live
AI development is outpacing our defenses. The researchers, engineers, and cryptographers working on closing that gap came to Berlin for our Secure & Sovereign AI Workshop and all talks are now live.
Explore ideas from 25+ speakers on AI security, privacy, and decentralized cooperation.
Highlights include:
Can automated red-teaming keep up with AI? Davide Crapis on automated red-teaming for AI security.
Can AI tell it’s being tested? Daniel Juhl on whether models know when they’re being evaluated.
Sovereign compute starts at home. Allison Duettmann on what it takes to own your own compute.
How do you evaluate security when AI agents interact? Ben Hagag on a framework for multi-agent security evaluations.
Where do privacy and cryptography meet AI? Janabel Xia on privacy and cryptography in AI.
Explore all talks on our YouTube channel.
More from Berlin: the past months at our AI Node
“I believe the Nodes, as a new type of third place, matter at a larger scale, and I believe they matter now more than ever.” – Bradley C Royes
What’s it like running an AI Node? In his second diary entry, Berlin Node Manager Bradley shares his reflections on a packed June-to-September at the Node: four grantee-led salons, the 72-person Secure & Sovereign AI Workshop, European Frontier AI Safety Day with SAIGE, and 50 researchers around a Thursday lunch table. Read it on Substack.
Bradley C RoyesNew Grantees
We’re excited to share our latest round of AI Node grantees: researchers and builders working on the technical foundations of safe, understandable, and cooperative AI, and AI-first science.
TINSEL: Open Biosafety and Post-Quantum Provenance Tracing for AI-Designed Biology
AI protein design can now produce thousands of novel sequences in a day, with no built-in way to verify what any of them are or where they came from. Babita Singh from Genethropic is building TINSEL, open infrastructure that lets protein-design tools check, cryptographically sign, and trace every sequence they generate, so each one shows whether it is safe, who designed it, and how it can be used.
Open Hyperbolic Embedding Models and Inspection Tools
Much of machine learning represents data in flat, Euclidean space. Matin Mahmood, founder of hyper³labs, studies how changing that geometry shapes what models learn. He trained Hyper3-CLIP, a hyperbolic vision-language embedding model for search that encodes hierarchy directly in its embedding space. With open models and inspection tools, he is investigating whether hyperbolic representations can make model behaviour easier to inspect and audit.
Automating connectomic proofreading for scalable brain mapping
Brain mapping is the path toward simulating minds, but every map contains millions of mistakes that currently need to be fixed by hand, a job that would take thousands of years for a single human brain. Tim Farkas, affiliated with the MIT Boyden Lab, is building AI to automate connectomic proofreading, making whole-brain mapping scalable for the first time.
Self-Driving Nanofabrication with Thinking Microscopes
Controlling the assembly of nanoscale building blocks remains a major challenge in nanofabrication. Vida Jamali from Georgia Tech is developing a programmable nanofabrication approach where agentic AI guides nanoscale assembly in real time: beam-powered particles act like tiny machines while an electron microscope observes their motion and adjusts its beam to steer and coordinate them into target structures.
Safeguarding Delegated AI Interaction
When we hand important tasks to AIs that act on our behalf, how can we make sure our preferences and goals are still represented, and that we don’t gradually lose control without noticing? Lara Kirfel, Max Planck Institute for Human Development, studies alignment and gradual disempowerment in human–AI interaction.
Closing the Inadvertent Leakage Channel
For AI assistants to be useful, they have to hold your private data: passwords, health records, financial details. Sanjam Garg from Carabid Inc. recently showed that today’s models leak these secrets through ordinary replies even while correctly refusing to reveal them, and no filter stops it. He is building the missing defense: provable, math-backed limits on what an AI’s outputs can leak.
Kinetic Inspection of Trajectory Embeddings (KITE)
Inside a language model, every piece of text is a trajectory moving through a high-dimensional latent space. Omar Ali Fdal and Matteo Giomi from Modal Resonance are using the geometric properties of those trajectories (velocity, curvature, shape) to build open-source monitoring that can catch prompt injection, jailbreaking, and anomalous behavior as it happens inside the model.
Glass.bio: data infrastructure for single-cell biology, with agentic analysis
Biology produces more data than scientists can actually use, and most of it is stuck in formats that take weeks of wrangling before analysis can even begin. Jonathan Hales and Florian De Rop from Glass.bio are building a searchable library of this data, starting with single-cell biology, paired with an AI assistant that helps researchers explore and analyze what they find.
Congratulations to this round of grantees.
Want to support our grants program? Reach out at partnerships[at]foresight[dot]org
We are currently seeking technical grant advisors with expertise in one of our focus areas.
Our grants support projects that use AI to advance science and safety across AI security, privacy, decentralized cooperation, human empowerment, neurotechnology, longevity biotechnology, and molecular nanotechnology.
Find out more and apply.
Allison Duettmann joins Foresight Board of Directors
After ten years leading Foresight Institute, Allison Duettmann is joining our Board of Directors as she transitions out of her role as President and CEO. Co-founder Christine Peterson will serve as interim President while the board searches for Foresight’s next CEO.
Under Allison’s leadership, Foresight became the organization you know today: six focus areas, a fellowship that has supported 266 researchers, a multi-million-dollar grants program, Vision Weekends on three continents, and AI for Science and Safety Nodes in San Francisco and Berlin, built on her conviction that open science needs open compute. And the team she built: 12 people across San Francisco and Berlin, running grants, fellowships, seminars, AI Nodes, Vision Weekends, and shaping what comes next.
We are very grateful to Allison for the past decade, and for everything still to come. We look forward to having Allison and the whole Foresight team at Vision Weekend next week in San Francisco!
Seminars
Watch the latest talks from our technical seminar groups. To attend future seminars live, request to join upcoming sessions on the Foresight calendar.
See all upcoming seminars
Community updates
PolyBio Convergence Summit | November 18–20, Cambridge, MA
Three days on the links between persistent infections, chronic disease, and aging, hosted at the American Academy of Arts and Sciences. Opens with a talk by Amy Proal (PolyBio Research Foundation; Foresight Health Extension Group), followed by sessions bringing together researchers from Gladstone, Karolinska, UCSF, and Sorbonne on viral lipid metabolism, cellular aging, and parasite-driven immunometabolic dysfunction. Complimentary registration codes available for the Foresight community — email [email protected] to request one.
2026 AE Global Summit on Open Problems for AI | London, Nov 25-27
Foresight Fellow Ruairidh Battleday’s organisation Thinking About Thinking has announced a new Summit on AI for Science: three days in London gathering researchers, engineers, founders, investors and policymakers with keynotes, panels, breakout rooms, an expo and a startup competition final. Details and tickets.
Aurelia Prize in Design for Space Urbanism | $100K
The Aurelia Institute’s 2027 global design competition is open for architectural concepts of lunar settlements — surface bases, orbital stations, rover fleets, mining outposts, or something no one has proposed yet. Jury includes NASA astronaut Victor Glover, architects Norman Foster and Gui Trotti, MoMA curator Paola Antonelli, and sci-fi author Adrian Tchaikovsky. Winner gets up to $100,000 and an invite to Aurelia’s Beyond the Cradle at MIT. Apply by January 15.
Biostasis startup Nectome: 2026 retrospective
If you’re following biostasis, Nectome has published a 2026 retrospective covering the year’s milestones: Lightcone Commons funding, a doubled team, a new preservation location in Oregon, and more. Read it here.Bylaw is hiring a Head of Research and a Founding Researcher (San Francisco, in person)
bylaw is an early-stage public benefit corporation working to ensure AI systems follow the law, building evaluations and post-training for law-following AI. Interested researchers can reach out to Morgan at [email protected].
© 2026 Foresight Institute
50 California Street, Suite 1500, San Francisco, CA 94111