Open-source AI backers share security concerns, urge provenance review

Policymakers should avoid creating policies that treat open-weight or open-source models differently from closed models and instead focus on tracking the origin and development of artificial intelligence technology more generally in order to address serious cybersecurity risks, according to advocates for a more open and interoperable AI stack.

“There's a push and a pull there. Everyone's worried there might be a proclamation that we can't use those,” IBM’s David Cox said referring to software emerging from China that is attractive...