Not all that glitters is gold. You’ve seen plenty of AI booster-ism from me in 2026, but the conversations I have with software companies also has a material slice of Governance/Risk/Compliance.
When experimenting with L3 and L4 AI tools, it’s not really negotiable whether we use sandboxes (or other principles of least privilege). If not, then we’re no longer discussing if you’ll have a material security incident, we’re discussing when.