Palo Alto Networks is using GPT-5.5 and OpenAI Codex in production security workflows, with engineer Gunjan Patel leading the implementation. The headline result: first-pass vulnerability reports are now detailed and actionable without manual cleanup.
Two specific gains stand out. GPT-5.5 runs tool calls in parallel rather than sequentially, which compresses multi-step security analysis into fewer tokens and less wall-clock time. That token efficiency is not a minor optimization, it changes the economics of running AI across an enterprise security team at scale.
The full video is worth watching for the workflow specifics: how Codex fits roles beyond engineering, and what parallel tool use actually looks like inside a real cybersecurity pipeline. The architecture decisions Patel's team made to get here are the part most summaries will skip.
[WATCH ON YOUTUBE →]