Airlines are getting hit right now. Air traffic control is getting hit. The NSA is getting hit. Banks and credit card companies are getting hit.
This is not a bad quarter. The attacker changed.
The attacker is now a harness: an autonomous loop that scans, probes, and chains identities around the clock. In November, the FBI issued a FLASH alert on ARTEX, an open source agentic pentesting system that has forked into dozens of variants. Cloudflare reports AI orchestrator attack traffic up 30x since October 1, with ARTEX and its forks at over 60% of known orchestrators. Researchers have counted more than 2,300 exposed offensive AI consoles on the public internet. State-aligned actors and criminal groups are already running these systems in live operations.
Most defense programs still run on a calendar built for a human attacker: annual penetration tests, quarterly scans, remediation queues measured in quarters. We have taken baby steps beyond preparing for a traditional attacker to attack us in a traditional way at a traditional speed.
Three shifts follow from that:
Finding vulnerabilities got cheap. Anthropic's Project Glasswing surfaced 10,000+ high and critical vulnerabilities across roughly 50 partners, and more than 99% of candidate findings sat unpatched. Remediation, not scanning, is the bottleneck.
The attack tooling is public, and defenders should run it against their own systems in a controlled way. Rehearse against the real attack, not a memory of one.
Defense has to become a loop. Scan everything with the best available models, fix with AI-assisted remediation and human review, then scan again. A fix counts only when the exploit path fails on a re-check.
I am writing a series on what defense looks like at machine speed: the harnesses, the theories that hold up, and the operating model underneath. Part 1, Baby Steps, is on Substack.