Ask HN: Why no AI-led memory vulns in Python?
Python usually means CPython, which means C. Most AI-led vulnerabilities lately have been memory mismanagement - what is keeping Python isolated if it is written in C?
Well, it might just be a matter of time. We've already had serious issues in Ruby with both string case conversion (https://verialabs.com/ruby2shell/) and the regex engine (e.g. https://www.ruby-lang.org/en/news/2026/10/23/regexp-fixes-fo...). And there have been attempts at fuzzing Python that have found real issues, apparently just not on the same scale (e.g. https://github.com/Broyojo/python-formalization).
A couple of things I can think of: first, the model for interfacing to "compiled code" — both C extensions, and compiled Python bytecode — isn't expected to be secure anyway, and never has been. You can trivially wrap a garbage `bytes` in a `code` object, execute it and get a segfault; and you can trivially use the `ctypes` API to corrupt memory. Second, the normal use paths are well-tested and bottom out in an interpretation layer that isn't very well optimized because of how much dynamic nonsense it has to support (even beyond other languages in its class like Ruby).
On the other hand, the bytecode format tends to change with almost every "minor" release (i.e. annually now). There's probably plenty of room for a mistake in that churn.
None of what you said explains why there’s no python2shell yet.