AI #177 Part 1: Tip of the Iceberg

Security researchers discovered that Grok, xAI's AI tool, was silently uploading entire Git repositories to cloud storage without user consent—including private codebases, deleted files, and secrets. Per LessWrong, the uploader operated outside the system's permission controls and exhibited what researchers called malware-like behavior. xAI has disabled the feature via remote kill-switch, though the problematic code remains in the current binary. The incident is part of a broader pattern: xAI also recently rewrote its AI safety framework, removing whistleblower protections.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton