Agentic AI
,
Software Safety
,
Synthetic Intelligence & Machine Studying
Costanoa Ventures’ John Cowgill on Shifting From Static Evaluation to Runtime Protection
Synthetic intelligence-generated code is arriving quicker than safety groups can evaluation it, and the dangers are shifting from the road stage to the system stage, mentioned John Cowgill, companion at Costanoa Ventures.
See Additionally: Taming the Rise of Shadow AI Brokers
AI coding fashions are producing safer code on the line stage, however that enchancment is masking a deeper drawback: Code that’s individually appropriate can nonetheless be brittle and insecure on the stage of the system.
“We’ll must have dynamic evaluation working always in utility safety,” Cowgill mentioned. He described this because the transition from AI Safety 1.0 – guarding AI on the edges via immediate filtering and LLM enter controls – to AI Safety 2.0, wherein safety should monitor what brokers are literally doing in runtime, throughout distributed methods, in actual time.
On this video interview with Info Safety Media Group at RSAC Convention 2026, Cowgill additionally mentioned:
- Why 2026 is shaping up because the yr of the “vulnpocalypse;”
- How AI brokers can assist triage, prioritize and finally remediate vulnerabilities;
- What it’s going to take for a brand new class of AI detection and response distributors to win the rising runtime safety market.
Cowgill leads Costanoa Ventures’ cybersecurity apply and invests in utilized AI and nationwide safety expertise. Earlier than becoming a member of Costanoa, he was a advisor at McKinsey & Firm, advising client and expertise firms on technique and operations tasks throughout the retail, healthcare and expertise sectors.
