Experts: human in the loop is now the bottleneck in AI security

At a roundtable in San Francisco, security experts agreed that AI systems have dramatically accelerated vulnerability discovery, but fixing them still depends on human approval. Participants noted an 8-to-1 ratio of issues found versus fixed, and predicted a sharp rise in security spending.

• Ratio of vulnerabilities found to fixed stands at 8 to 1
• Humans must approve every fix, slowing remediation
• Ed Sim of Boldstart Ventures predicts a surge in security spending
• AI agents already handle vulnerability triage well