Verifiable Compute Foundation
Containing the most serious risks from advanced AI will likely require agreement between companies and governments, and any such agreement is only as good as each party's ability to verify that the others are abiding by it. An emerging toolbox of AI verification techniques, spanning hardware-based mechanisms and cryptographic approaches, could give governments a way to verify claims about AI development and use while protecting models, data, and trade secrets. The Verifiable Compute Foundation is an independent nonprofit working to get these techniques ready for the moment they're needed, and serving as a partner to governments and other stakeholders along the way. Seismic monitoring helped make nuclear test bans verifiable. The equivalent for AI remains to be built. The cluster is open to academic labs, nonprofits, startups, industry and government teams doing serious work on AI verification. It is operated for VCF by Lucid Computing.