Tokyo · Sunday 4 October 2026

Turn an idea into a funded AI safety project

A one-day, in-person workshop for people in Japan who are already engaged with AI safety and want to do it full-time. We work through theories of change, write our ideas down, red-team them, and leave with a grant application worth submitting.

Register your interest →
Date
Sun 4 Oct 2026
One full day, in person
Location
Tokyo
Venue to be confirmed
Places
About 20
Small enough to workshop everyone's idea
Cost
Free
Meals provided on the day
Why this, why here

Japan has the people. It does not yet have the jobs.

There is a large pool of technically skilled engineers and researchers in Japan who could be strong contributors to technical AI safety research — and a very small number of positions for them to move into.

107
volunteers across nine AI safety organisations in Japan
2.5
funded full-time-equivalent positions at those same organisations
~20
people we can work with closely in a single day

Those first two figures come from a recent analysis of AI safety funding in Japan. The gap they describe is an opportunity: a lot of people are already doing this work part-time, alongside demanding jobs. What stands between them and a funded role is usually not talent or motivation. It is a crisp theory of change, an idea that survives contact with a critic, and a grant application that actually gets written and sent.

That is what this day is for.

The day

Discuss, write, red-team, submit

This is a working session, not a lecture series. Expect to spend most of the day either arguing about your idea or writing it down.

01
Theories of change
How research actually reduces risk — and how to tell whether yours does. We use the misuse / misalignment / systemic risk framing to get precise about which problem a project is really attacking.
02
Choosing a problem
What makes a research problem worth your next year: tractability, neglectedness, and the difference between a topic and a question. Iteration on your own idea with someone who has done this professionally.
03
Write it down
Everyone commits their idea to paper. Vague plans stay vague until you have to write them in full sentences; this is the part most people skip.
04
Red-team each other
Structured criticism in small groups. The goal is to find the weakest assumption in the room's ideas before a grant reviewer does.
05
Submit, or keep iterating
We map your idea onto the open funding calls that fit it, and you leave with a draft to send. Anyone who wants to keep refining continues online afterwards.
Who it's for

Come with something you want to work on

The emphasis is on technical research, but governance and field-building projects are equally welcome — the reasoning is the same, only the outputs differ.

Engineers and researchers
You have the technical background and follow the field, but you are doing safety work in evenings and weekends. You want to know what a full-time version of it would look like.
Students and academics
You are choosing a thesis topic or a next project and want it to matter. Bring the shortlist you are deciding between.
Community organisers
You run or help run a group and want funding for field-building work. Same exercise: what changes because you did this, and how would you know?
Newer to the field
You do not need a track record in AI safety. You do need a real interest and a willingness to have your idea taken apart in a friendly way.

The workshop runs in English.

Who's running it

Facilitator

David Williams-King
David Williams-King

David works on getting people onto projects that matter for existential risk — it is his day job, and it is what he does with most of the rest of his time. He has been lead instructor at AI Safety Bootcamp Singapore and at FAST (Frontier AI Security Training), has taught this material to the M3 fellowship in London, and writes The Safety Apprentice, a guide to getting into the field.

Most relevant here: Choosing Research Problems, which is roughly where this workshop starts.

Attending

Register your interest

Places are limited to about twenty so that every attendee's idea gets real attention. The workshop is free to attend and meals are provided.

The venue is being finalised — it will be in central Tokyo, and everyone who registers will be told the location and timings as soon as they are set.

Venue and registration link to be announced