OpenAI President Greg Brockman tested the security of his static website—hosted on AWS behind Cloudflare—using ChatGPT. In just 15 minutes, the model identified 13 configuration issues. The list of vulnerabilities included missing SPF and DMARC records for domain email, an outdated jQuery library riddled with known exploits, and unencrypted HTTP traffic between Cloudflare and the AWS origin server. As Brockman admitted, he barely understood the nuances of some of these settings himself, yet critical misconfigurations had lingered in his infrastructure for years.
Brockman then leveraged the model's agentic capabilities to patch the discovered flaws. The AI agent independently accessed the Cloudflare management console, configured the correct DNS records, enabled end-to-end encryption, removed the vulnerable jQuery script, rolled out DMARC step-by-step, and migrated the entire project from AWS to Cloudflare Pages without manual intervention. The entire remediation process took exactly one hour.
For business leaders, this experiment signals a tectonic shift: entry-level security audits and routine vulnerability remediation are dropping out of high-priced consultancy scopes. Basic cyber hygiene is quickly becoming an automated script that a CTO or product owner can run over a lunch break, cutting out five-figure advisory retainers.