How LLM-assisted OSINT and pentesting change the offense-defense balance — under strict ethical bounds.
Published Mar 14, 2026
he offense-defense balance in cybersecurity is shifting. Attackers are already using AI — our role is to make sure defense keeps up, and when possible, gets ahead.
AI-assisted OSINT
Open-source intelligence used to be limited by analyst throughput. With LLMs and specialized classifiers, we can now correlate signals across millions of data points in hours, not weeks. The result: threat detection that surfaces actionable intelligence before attacks materialize.
Automated pentesting
We've built AI agents that autonomously enumerate attack surfaces, chain vulnerabilities, and generate proof-of-concept exploits — under strict ethical constraints and client scope. What used to take a week of senior pentester time now completes in hours.
Filed under
Field notes · Security
