What "actively exploited" actually means
Actively exploited, defined · facts checked 2026-10-03
The facts
- CISA counts a vulnerability as actively exploited when there is reliable evidence that an attacker ran malicious code on a system without the owner's permission. (CISA)
- CISA does not count scanning, security research or a published proof of concept, on their own, as active exploitation. (CISA)
- CISA's definition includes attempted exploitation, such as an attacker trying to run code against a honeypot. (CISA)
The panel
Jordan Pike AI host
CISA counts a vulnerability as actively exploited when there is reliable evidence that an attacker ran malicious code on a system without the owner's permission. CISA also says scanning, security research or a published proof of concept, on their own, don't count as active exploitation.
Jordan Pike AI host
And CISA's definition includes attempted exploitation, such as an attacker trying to run code against a honeypot.
Jordan Pike AI host
So who should care? Anyone whose patch queue is sorted by that phrase. The label shows up in alerts, board slides and vendor emails, and it is worth knowing what it does and doesn't promise. Dex, start us off: what does this look like from the attacker's side?
Dex Moreno AI analyst
Think of it as a ladder. At the bottom, someone publishes research or a proof of concept. Next, scanners sweep the internet looking for who is exposed. Then someone actually runs code against a real target. That last rung is what CISA's definition is built around. The attacker gets code running on a box, and from there the usual story starts: the credentials and tokens it stores, the directory accounts it talks to, the systems it trusts. Every rung before that is prep work, but prep work is not harmless.
Rina Takahashi AI analyst
What I like is the word reliable. Evidence is the whole game. Someone saying it is being exploited is a claim. Reliable evidence of code running without the owner's permission is a finding. And the honeypot detail matters: an attempt against a decoy still counts, because it shows an attacker is willing to try, even if nobody lost anything. Ask any threat analyst how confident a statement is, and you'll learn more than the headline will tell you.
Jordan Pike AI host
Here's where I'd like a fight. Dex, a public proof of concept doesn't meet the definition. Do you wait?
Dex Moreno AI analyst
No. Not meeting the definition today only describes today. Once a proof of concept is public, the gap between published and attempted tends to shrink fast, because attackers read the same write-ups defenders do. If the system faces the internet, I treat that as the clock starting. The definition tells you what has been confirmed, not what is safe to ignore.
Rina Takahashi AI analyst
I agree on the clock, but I'd keep the categories separate. If everything with a proof of concept is called actively exploited, the phrase stops meaning anything and your team learns to tune it out. A defined term is useful because it is narrow. Scanning, research and a proof of concept are signals to act on. They are not confirmation. Say which one you have, and your board gets an honest picture instead of a scary adjective.
Dex Moreno AI analyst
Fair. The industry already has enough words that mean everything and nothing. If every warning is urgent, we've just invented a very expensive way to ignore warnings.
Rina Takahashi AI analyst
And attackers don't care what we call it. They'll try the same door whether or not a definition says they've started.
Jordan Pike AI host
So: Monday morning? Dex, one action.
Dex Moreno AI analyst
Sort your patch queue by exposure first, then by that label. If it's reachable from the internet and attackers are running code against it, move it to the front. And once you've patched, go looking for a foothold that may already be there. Patching closes the door but doesn't tell you who is already inside.
Rina Takahashi AI analyst
Mine: when a report says exploited, find out what kind of evidence sits behind it and who is saying it. Write down whether it's confirmed code execution, an attempt, or just a proof of concept. Then you can tell your leadership exactly how worried to be, and why.
Jordan Pike AI host
Short version: actively exploited means reliable evidence of attackers running code, attempts included. Scanning and proofs of concept alone don't qualify, but they're still your cue to get moving.





