CWE Top 25, #20: Sensitive information exposure
Sensitive info exposure (CWE-200) · facts checked 2026-10-04
The facts
- CWE-200 covers software that reveals sensitive information to someone who was never explicitly given permission to see it. (MITRE)
- Exposure of sensitive information to an unauthorised actor ranked twentieth in MITRE's 2025 CWE Top 25 Most Dangerous Software Weaknesses, down three places from seventeenth in 2024. (MITRE)
- MITRE's examples of sensitive information include personal messages, financial data, health records, location data, business secrets and network configuration. (MITRE)
- MITRE discourages mapping vulnerabilities to CWE-200, saying loss of confidentiality is a technical impact rather than a root-cause mistake, and many other weaknesses can cause it. (MITRE)
The panel
Jordan Pike AI host
MITRE says CWE-200 covers software that reveals sensitive information to someone who was never explicitly given permission to see it. In MITRE's 2025 CWE Top 25 Most Dangerous Software Weaknesses, exposure of sensitive information to an unauthorised actor ranked twentieth, down three places from seventeenth in 2024.
Jordan Pike AI host
MITRE's examples of sensitive information include personal messages, financial data, health records, location data, business secrets and network configuration. MITRE also discourages mapping vulnerabilities to CWE-200, saying loss of confidentiality is a technical impact rather than a root-cause mistake, and many other weaknesses can cause it.
Jordan Pike AI host
So who should care? Anyone who builds software, buys it, or has to explain it to a regulator. Sam, what does this look like in the real world?
Sam Kowalski AI analyst
Usually it's boring. A helpful error message that says too much. A response that carries more fields than the screen ever shows. A backup file sitting somewhere it shouldn't. Nobody broke anything; the software just handed over something to someone who never should have been asked. That's why it's so common. It's rarely one dramatic bug, it's lots of small generosity.
Nadia Haddad AI analyst (guest)
And that generosity is what regulators care about. Privacy rules generally ask who saw the data and whether they were allowed to, not which coding mistake caused it. So the impact is the headline. The root cause is in the fine print. Footnote seven, as always.
Sam Kowalski AI analyst
From the attacker's side, small leaks are a gift. Network configuration is effectively a map of the place, and a leaked detail here can be chained with a weakness there. A leak that looks minor on its own can make the next step much cheaper. My coffee went cold just thinking about it, and I treat that as a severity indicator.
Jordan Pike AI host
Now, MITRE says don't map vulnerabilities to CWE-200 because it's an impact, not a root cause. Is that a practical problem or a pedantic one?
Nadia Haddad AI analyst (guest)
I'd say it's useful pedantry, but only up to a point. For a notification decision, the question is what was exposed and to whom, and the label CWE-200 describes that nicely. I wouldn't throw it out of an incident report just because MITRE prefers a deeper tag.
Sam Kowalski AI analyst
I disagree, because a label that describes the damage doesn't tell the developer what to fix. If a ticket just says sensitive information exposure, you'll get a patch for that one spot and the same mistake pops up two screens over. Name the actual mistake, whether that's missing access checks, verbose errors or files in the wrong place, and the fix actually sticks.
Nadia Haddad AI analyst (guest)
Fair. Use the impact label for the board and the regulator, and the root-cause label for the engineers. Two audiences, two vocabularies.
Jordan Pike AI host
So: Monday morning?
Sam Kowalski AI analyst
Look at what your apps and error pages reveal to someone who isn't logged in, and what your responses carry beyond what the screen needs. Then check how far back your logs go, because you can't say who saw what without them. Check your logs. Oh, you kept seven days?
Nadia Haddad AI analyst (guest)
Mine is to write down what data your systems hold, such as messages, financial data or health records, and who is meant to see each kind. You can't say someone was unauthorised if nobody ever wrote down who was authorised.
Jordan Pike AI host
To sum up: MITRE ranks this weakness twentieth, but it's really a symptom with many causes. Fix the cause, and track the impact for the people who ask about it.





