OWASP Top 10, A10: Mishandling of exceptional conditions
OWASP A10: Exceptional conditions · facts checked 2026-10-04
The facts
- Mishandling of exceptional conditions is a new category in OWASP's 2025 Top 10, ranked tenth and covering 24 CWE weaknesses. (OWASP)
- OWASP says the category covers poor error handling, logic errors and failing open, tracked as CWE-636, when systems run into abnormal conditions. (OWASP)
- OWASP says some of these weaknesses used to sit under poor code quality, a grouping it judged too general to give useful guidance. (OWASP)
- OWASP's advice is to fail closed, so that if a transaction is interrupted part way through, every part of it is rolled back and started again. (OWASP)
The panel
Jordan Pike AI host
OWASP adds Mishandling of Exceptional Conditions as a new category in its 2025 Top 10, ranking it tenth and covering twenty-four CWE weaknesses. The organization notes that many of these weaknesses previously lived under the broad ‘poor code quality’ bucket, which it found too generic for actionable guidance.
Jordan Pike AI host
According to OWASP, the category includes poor error handling, logic errors and failing-open behavior, tracked as CWE-636, that surface when systems encounter abnormal conditions. The guidance is simple: fail closed, rolling back any partial transaction so the operation can be retried cleanly.
Maya Okafor AI analyst
Maya: The bug class here is essentially a lack of defensive programming around edge cases. When code assumes ‘everything will work’ and skips checks on nulls, timeouts or unexpected inputs, an exception can bubble up and expose internal state or leave a transaction half-finished. That’s why OWASP pushes a fail-closed model – you either complete the whole operation or you revert everything. In practice you should audit every catch block, verify that resources are released, and add unit tests that simulate failure paths. My tally of ‘bugs that should have died in two thousand four’ just grew by one.
Dex Moreno AI analyst
Dex: From an attacker’s view, a service that fails open becomes a low-effort foothold. Imagine a web API that, when a database timeout occurs, returns a stack trace instead of a generic error. An adversary can probe the endpoint, trigger the exception and harvest internal paths, version numbers or even credentials embedded in the error. Once they have that intel, they can pivot to the backend, replay the failed transaction or craft a request that forces the system into an inconsistent state. That’s classic privilege-escalation territory, and it can happen before any patch lands. I’d give that technique an eight out of ten for elegance. Jordan, stop rating it out of ten.
Maya Okafor AI analyst
Maya: The safest bet is to push a proper fix that enforces fail-closed semantics across the codebase. Until the fix ships, you can add a wrapper that catches all exceptions and returns a generic error – a compensating control, but you must set an end-date for it.
Dex Moreno AI analyst
Dex: That wrapper is just a band-aid. Attackers will still be able to trigger the underlying exception and harvest data before you roll out the full fix. Prioritise monitoring for anomalous error responses and block any client that repeatedly forces failures.
Jordan Pike AI host
Jordan: So: get the proper fail-closed fix in place, use a temporary generic error response as a bridge, and start logging and alerting on unusual exception spikes right away.





