Skip to content
Security & Trust Talking point

Defenders should spend on shrinking exposure and containment, not on patching alone

A Black Hat panel argued patching cannot keep pace with AI-found flaws, so budgets should favour exposure and containment.

W
WebPulse Newsroom
AI-assisted · 2 min read
Share on X LinkedIn
Defenders should spend on shrinking exposure and containment, not on patching alone
In brief
  • Katie Moussouris and Morgan Adamski argued at Black Hat that AI-found flaws will outrun patching, so budgets should favour less exposure, segmentation and containment.
  • Adamski said a breach is likely and should be planned for in the budget. Moussouris said patching still deserves investment, just not first.

At Black Hat USA 2026, a panel on AI and cyber defence argued that patching cannot keep pace with AI-found flaws. Katie Moussouris, founder and CEO of Luta Security, and Morgan Adamski of PwC said money should go to shrinking what is exposed, splitting networks into zones and limiting damage. Adamski added that a breach is likely, so budgets should plan for one.

What was said

Moussouris said patching never scaled, even before AI, because patches cost money to make and deploy. Now AI is finding issues in far greater volume and at greater speed, she said.

She said: "No, we could never patch our way out of this." She also criticised ranking patches by CVSS scores, a public severity rating. In her view that ranking comes from outside and ignores your own protections. AI can also chain several flaws together, she said, which ended any illusion of staying ahead.

Patch management is still a place to invest, Moussouris said, but not the biggest return. She put first the reduction of what is exposed to the internet, then segmentation. Adamski said the assumption of a fix for every flaw is wrong. In her words, "there is not going to be potentially a patch for every vulnerability that's discovered via AI." She urged firms to know their crown jewels, meaning the assets they cannot lose, so they can contain an intruder and recover.

Why it matters

Our reading: the panel's argument comes down to one idea. If AI-found flaws arrive faster than teams can fix them, the control that still works is the one that limits what a flaw can reach.

The mechanism is simple. A severity score rates a flaw alone, yet an attacker can chain small flaws into a serious break-in. An outsider cannot try a flaw on a system the internet cannot reach. Segmentation adds walls between systems, so an intruder in one zone cannot freely walk into the others.

A budget owner can therefore ask three things: what is exposed to the internet, where are the crown jewels, and how would we contain and recover?

The other side

The panel did not say patching is dead. Moussouris still called it a place to invest. The moderator also asked how to win over firms that never embraced the basics. Adamski answered by framing the case as business risk, such as lost sales during an outage.

Segmentation also depends on knowing what you own. Sergiy Konovalov of Anthropic said the industry average is about two in ten operators who firmly understand their assets. The excerpt does not say which operators he meant. It is one speaker's rough figure, taken from a machine transcript.

The panel offered no figures on what containment costs or how much damage it prevents.

Written by the WebPulse Newsroom with AI assistance, and checked by our editorial review: every quotation was verified against the recording's transcript. How we use AI.

The conversation this talking point comes from