Security research at Apple just got an AI assist.
Under a project called Glasswing, Apple is working with Anthropic to use Claude for finding vulnerabilities in iOS, macOS, and Safari before attackers do. Details are limited, but the arrangement involves Claude helping Apple's security teams identify bugs and weaknesses through automated code analysis - scanning for patterns that could be exploited, rather than waiting for researchers or adversaries to find them first.
Why Apple Would Bring In an Outside AI Vendor
Apple runs software on over a billion active devices. iOS zero-days - previously unknown security flaws that attackers can exploit before a patch exists - are worth millions of dollars on exploit markets. The company runs a substantial bug bounty program and employs large internal security teams, but the volume of code across three operating systems and a browser is enormous. Automated AI assistance for vulnerability scanning is a logical addition to that infrastructure.
What's notable is that Apple chose an external vendor at all. The company has historically been protective of its security operations. Security research on Apple's own code, though, doesn't involve user data - so the privacy sensitivities that typically make Apple cautious about external AI partnerships don't apply the same way here.
For Anthropic, Glasswing adds to a pattern of high-value technical partnerships in security and government work, where Claude is being applied to code analysis and reasoning tasks rather than content generation.
Apple hasn't disclosed the scope of the collaboration or whether it extends beyond the three named platforms.