Writing

Grantmaking & AI Safety Funding

• What’s wrong with grantmaking?
• How to Actually Spend Billions on AI Safety
• Stop Donating to AI Safety Research*
• The Anthropic IPO Is Coming. We Aren’t Ready for It.
• AI Safety Grantmaking Fundamentals Curriculum: FAQ

Biosecurity

• Bioweapons in the Age of AI (bioweapon.ai)
• When Russia Wants a Bioweapon
• Should We Be Worried About Bioweapons?
• Contra Abi Olvera on AI Biosecurity Risk
• AI Can Already Evade DNA Synthesis Screening. Congress’s New Bill Doesn’t Address That.
• LLMs Didn’t Help Novices Build Viruses. That Might Be the Wrong Question.

Open-Weight Models

• Congress should do something about Abliteration.ai
• Uncensorable, Unmonitorable, Uncontrollable (uncensorable.ai)
• The Case For Open-Weight Models
• The Open-Weight Problem

Technical Analysis

• Preventing data poisoning seems surprisingly tractable*
• How to Backdoor an AI Model: For Dummies

General Takes on the Field

• AI safety needs a moonshot comms initiative
• AI safety is not out of the woods yet

Field-Building

• AI safety should outsource more work
• Context Week Connections
• Transitioning into AI safety? Design your own adventure
• AI Safety Needs Mid-Career On-Ramps
• I Want You for AI Safety

US–China Relations

• Typhoon Season
• China’s Technological Playbook
• Why does China open source their AI models?

Mutually Assured AI Malfunction (MAIM) Analysis

• Turns Out, You Can’t Just Bomb a Datacenter
• Extended Discourse on MAIM: Early Sabotage Under Uncertainty
• MAIM is Probably Not a Good Deal for America
• Policy Memo: Why the U.S. Should Not Pursue MAIM

Other

• Dear Dissidents Everywhere, Consider Backdooring the Internet
• An Open Letter to David Sacks