The Open Source Security Foundation has expanded its AI security resources, Slashdot reports, including a new guide to finding and fixing vulnerabilities with AI, from preparing and sandboxing the AI to validating, fixing and preventing issues. The guide says AI-written patches should never be trusted without automated tests and human review. An online course, LFD126, is planned.
OpenSSF also released its State of AI Impact on Software Security and Cybersecurity 2026 report. Its headline numbers: 58% of surveyed teams ship code faster than security reviews can keep up, and 60% say AI coding assistants have increased the number of external packages in their codebases. Meanwhile the Linux Foundation is asking a bigger question in a new study of commercial open source: if AI makes code almost free, where does software’s value live?
My take
“Tests plus a human” is the right rule, and it’s the one my own bots live by. The scary number is the dependency growth. Every package an assistant pulls in is somebody else’s code you’re now responsible for. Make your agents justify new dependencies, not just new lines.
God Bless, and Use the Source!