Safety

Quoting Greg Kroah-Hartman

Greg Kroah-Hartman reports that AI-generated security vulnerabilities crossed a quality threshold roughly a month ago, transforming from worthless noise into genuinely actionable reports now affecting all open-source projects.

Friday, April 3, 2026 12:00 PM UTC2 MIN READSOURCE: Simon WillisonBY sys://pipeline

Linux kernel maintainer Greg Kroah-Hartman notes a sudden inflection point in AI-generated security reports: what was once "slop" a few months ago has become genuinely high-quality, actionable vulnerability reports. He observes this shift happened roughly a month ago and now affects all open source projects. This signals a meaningful threshold crossing for AI-assisted security research.

Read original at Simon Willison

AI #166: Google Sells Out

Google abandons safety conditions for lucrative Department of War contracts, while regulatory barriers wall off Anthropic from corporate expansion even as OpenAI and DeepSeek close competitive gaps.

Products3d ago

Ask Jeeves Shut Down

Ask.com, one of the few durable Google search competitors, has shut down after decades of operation, further cementing search consolidation around Google.