- Morgan Adamski called timely context on attacker intent extremely valuable, because it helps defenders choose which flaws to fix first.
- She also asked for openness on which AI cyber models are covered and which are not, and said old, undated indicators are no help.
Morgan Adamski, a PwC cyber leader who has left US government service, spoke on the Black Hat USA 2026 panel on AI and cyber defense. She said government can add real value by sharing timely context on what attackers intend. She also wanted openness about which AI cyber models are covered and which are not. Old, undated indicators, she said, are not what that sharing should turn into.
What was said
Adamski said she no longer speaks for the government. She offered a historical view of where sharing should go and where she thinks it is now. Intelligence agencies, she said, hold unique insight into the intent of nation-state and criminal attackers. That includes which weaknesses they are building tools to exploit.
Shared in real time, she said, that insight is extremely valuable. It helps CIOs and CISOs decide which vulnerabilities to patch first. She said some of this is already happening in real time because of AI. It has to happen even faster, she said, and rest on trust between the two sides.
She said useful sharing does not mean indicators that are three months old and carry no date. Those, she said, are "irrelevant and have no context". The conversation must explain why something matters.
Her second point was transparency about covered models, and about the review the executive order requires. Some things must stay internal. Still, she wanted people to understand "what's available and what's not" for cyber defense.
How intelligence reaches a patch queue
An indicator is a clue that marks known attack activity, such as a file fingerprint or a web address. A bare list says what was seen, not why it matters.
Teams face a long list of vulnerabilities, meaning known software flaws, and cannot fix them all at once. If they learn that a named adversary is actively exploiting one flaw, it moves ahead of the rest. A bare list cannot say that, so no flaw on it has a reason to come first.
Why it matters
Our reading: when not every flaw can be fixed, choosing the order is the real work. Adamski said on the same panel that AI-discovered vulnerabilities may not each get a patch.
A wrong choice is paid for by an organisation with a small security team, and by its customers. An undated feed with no context hands that choice back to them.
Buyers and managers can test any intelligence source. Does each item carry a date and a reason it matters? Do your defenders know which AI models they may use?
The other side
Adamski spoke as a former official, not for the government. She said some real-time sharing already happens, but the excerpt gives no example and no sense of how often.
She conceded that some information must stay internal. She said useful sharing depends on trust. The excerpt does not say how that trust is built, or how fast.
It also leaves open what the executive-order review will decide about covered models. Until then, defenders cannot be sure which tools they can count on.
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
- Black Hat: Black Hat USA 2026 | AI and the Future of Cyber Defense Panel (2026-11-13)





