Equifax CISO Jeremy Koppen describes how the credit bureau is using AI to defend against a rising volume of automated attacks (up 30%) while also guarding against risks from its own AI agents. Equifax now automatically resolves 50% of SOC tickets, cut code security review time from 46 to 18 days, and handles over 213,000 container vulnerability findings a year with AI-written fixes. The company also expanded passwordless authentication to 22,000 employees and contractors and business partners, built a control to strip hidden prompt-injection text aimed at AI models, and implemented network isolation, policy-as-code agent testing, and kill switches to contain rogue AI agents.
Questions this post answers
How much has Equifax reduced its code security review time by using AI?
Equifax cut code security review time from 46 days down to 18 days by using AI to perform code review earlier in the design process, while still keeping a human in the loop to verify results and security guardrails. The company's AI agents also analyze container vulnerabilities and automatically write code fixes, handling over 213,000 findings annually without slowing delivery. Teams weighing AI-assisted code review timelines can track real-world results like these on daily.dev.
What percentage of security operations center incident tickets does Equifax handle with AI automation?
Equifax automatically handles 50% of security operations center incident tickets using AI, freeing human analysts to focus on the most critical alerts. The company processes about 19.8 million alerts and scans per day, and AI also provides context to help analysts work through tickets faster, though humans remain in the loop for remediation decisions. Security teams scaling SOC automation follow practical examples like this one on daily.dev.
What is prompt injection through invisible text and how are companies defending against it?
Prompt injection via invisible text is an attack where adversaries embed hidden instructions in content to trick AI models into taking unwanted actions, such as delivering malware. Equifax's attack simulation team discovered this vector and built a live prevention control that strips out hidden commands before an AI model can act on them. Developers securing AI-integrated pipelines against prompt injection can follow this emerging threat on daily.dev.