North Korean Hackers Used AI Deepfakes to Compromise Axios NPM Package
What Happened
North Korean threat group UNC1069 used AI-generated deepfakes of company executives in fake Slack workspaces and Teams meetings to socially engineer maintainers of top npm packages. The axios HTTP library — present in an estimated 80% of cloud environments with 100 million weekly downloads — was briefly compromised with a cross-platform RAT. The malicious versions were live for approximately three hours before removal, but Wiz estimates 3% of scanned environments were affected. Maintainers of lodash, dotenv, fastify, and Node.js core packages were also targeted.
My Take
This is the supply chain attack that the security community has been warning about, except now the social engineering layer runs on AI. The attackers did not find a zero-day. They did not brute-force credentials. They cloned a founder's face and voice, built a convincing Slack workspace, and got a human to trust them. Three hours was enough to hit 3% of cloud environments. That math is terrifying. For anyone vibe-coding a production app and running `npm install` without thinking twice: this is your wake-up call. The speed at which AI lets you build is the same speed at which AI lets attackers operate. Every dependency you pull in is a trust decision, and most of us are making hundreds of those decisions on autopilot. The real shift here is that social engineering now scales. A deepfake meeting with a single maintainer can compromise millions of downstream applications. The open-source trust model was built for a world where attacks required human effort per target. That world is gone.
Read Original Source