The Great Refactor | IFP

How to secure critical open-source code against memory safety exploits by automating code hardening at scale

This essay is part of The Launch Sequence, a collection of concrete, ambitious ideas to prepare the world for advanced AI. These projects need people to build them. Get in touch. Summary Critical systems across the country run on software riddled with vulnerabilities. AI now arms our adversaries with automated tools to find and weaponize these flaws. But it is also a transformative opportunity to strengthen our defenses. A major source of vulnerabilities in our digital infrastructure stems from code written in languages like C and C++, which are vulnerable to exploits targeting computer…

saved by

• Oscar Marret

related reading

• Project Glasswing: Securing critical software for the AI era \ Anthropicanthropic.com
• Vulnerability Research Is Cooked - Quarrelsomesockpuppet.org
• AI for Program Reasoningai4pr.github.io
• When AI Writes the World's Software, Who Verifies It? — Leonardo de Mouraleodemoura.github.io
• Ten AI safety projects I'd like people to work onthirdthing.ai
• Your Solution Doesn't Know Your Problem Existsblog.atlascomputing.org
• Lean Software Scaling Laws · Gwern.netgwern.net
• Announcing Safety Research Grantsthinkingmachines.ai
• How Could Formal Verification Help with Cyber Resilience?astrangeattractor.substack.com
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• Eliminating Memory Safety Vulnerabilities Once and For Alldarpa.mil
• Eliminating Memory Safety Vulnerabilities Once and For Alldarpa.mil