ACCORDING to Dark Reading, security chiefs are bullish on AI and are planning to deploy more tools across their organisations, with Reddit CISO Frederick Lee and Omdia’s Dave Gruber weighing in on real‑world use. They describe how AI and large language models are being used to automate playbooks, convert run books into agents, and turn conversational input into queries for BigQuery or Splunk, expanding coverage and speeding analyst responses.
They also emphasise vertical use cases such as threat intelligence analysis to speed up operationalisation and reduce risk, while noting the rapid adoption cycle driven by clear value. The discussion also covers risks, including prompt injection, data hygiene, and governance, with calls for a structured architecture, an MCP gateway, and guardrails to keep humans in the loop.
A Veracode study mentioned in the dialogue found that AI models chose secure implementations only 45% of the time and failed to prevent cross‑site scripting 86% of the time, underscoring the need for ongoing human audit and secure coding practices. The speakers stress that success comes from starting narrow, defining guardrails, and collaborating with vendors and service providers to capitalise on AI’s potential while managing costs and risk.