A weekly cybersecurity news roundup covering several major stories. The main story involves a Vercel data breach traced back to a compromised OAuth token from a Context.ai security incident, where an employee had granted excessive Google Workspace permissions to a third-party app. The host argues this was fundamentally an OAuth hygiene failure and shadow IT problem, not an AI-driven attack as Vercel's CEO framed it. Other stories include: Claude Opus (referred to as 'Mythos') achieving 73% success on expert-level CTF tasks and completing a 32-step corporate network attack simulation; NIST announcing it will deprioritize CVE enrichment in the NVD due to a 263% surge in submissions since 2020; Cal.com moving from open source to closed source citing AI security concerns; and brief mentions of Zoom's AI detection feature, Trail of Bits beating Google's ZK proof, and Meta partnering with PortSwigger for Burp Suite licenses.
Sort: