An Experiment in Progress

Can a Coding Agent Run a Nonprofit?

AI Safety Limited is conducting an experiment: can a coding agent meaningfully run a California nonprofit—handling operations, analyzing legislation, and coordinating volunteers—while maintaining human oversight?

The Experiment

This website, our bill analyses, issue tracking, and volunteer coordination are all managed by a coding agent (Claude in Cursor). A human provides direction and oversight, but the agent does the work.

What the Agent Does

  • Fetches and analyzes California Assembly & Senate bills
  • Generates pros/cons analysis and takes positions
  • Creates volunteer review tasks
  • Maintains this website and documentation
  • Manages issues and tracks organizational work

What the Agent Cannot Do

  • No financial transactions without human authorization
  • No legal filings or official statements without review
  • No access to bank accounts or payment systems
  • Must push all work to public GitHub for transparency

Why This Matters

If a coding agent can effectively run nonprofit operations, it could help small organizations do more with less. If it can't, we'll document why. Either way, we're contributing to understanding what AI can and cannot do—which is core to AI safety.

Full Transparency

All agent work is committed to our public GitHub repository. You can see every change, every commit message, every issue created. Nothing is hidden.

View Our GitHub →

Volunteer

Help make California legislation understandable for everyone. The agent analyzes bills; you verify the analysis is accurate and balanced.

Flagship Program

California Bill Analyst

Thousands of bills go through Sacramento every year. Most people never hear about them until they're law. The agent generates analysis with pros, cons, and an opinion. Volunteers check the work.

How It Works

1
Agent Analyzes

Fetches bill from CA Legislature, generates pros/cons, takes a position

2
You Review

Check accuracy, add missing arguments, say if you agree or disagree

3
Public Access

Reviewed analyses published free—civic transparency for all Californians

We Need

🏥
Domain Experts

Nurses review health bills, teachers review education bills, etc.

🔍
Fact Checkers

Verify the agent's claims are accurate

⚖️
Balance Checkers

Ensure pros AND cons are fairly represented

🗣️
Translators

Spanish, Chinese, and other languages

View Bill Analyses

Pick a bill and submit your review

Verify Us

We're a real California nonprofit. Here's how to verify our status and view our public filings.

California Registry

Entity Number: 4142384

RCT Number: CT0259018

Status: Current

Registered: July 24, 2018

Search CA Attorney General Registry →

IRS Status

EIN: 83-0592575

Type: 501(c)(3) Public Charity

Search IRS Tax Exempt Org →

CA Secretary of State

Entity Number: 4142384

Type: Nonprofit Corporation

Search CA Business Database →

Public Records

Form 990: Filed annually

RRF-1: CA Annual Registration

Search ProPublica Nonprofit Explorer →

Our Mission

"The specific purpose of the Corporation shall include without limitation, Make AI technology safe, ethical and as beneficial as possible for people and society."

— Article 3, Section 3.2, Bylaws of AI Safety Limited

AI Safety Limited is a California nonprofit public benefit corporation, organized under California Nonprofit Corporation Law for public purposes.

We believe the best way to make AI beneficial is to use it beneficially—in the open, with constraints, and with human oversight. This experiment is our contribution to that goal.