News for hackers. Stuff that matters.
Re:Claude Code and Codex reverse engineered my 1994 Jeep's E... by nullc on Monday November 16, 2026 @11:33AM (#50519497)
Not just trained, but since you can only use them via an API they can have an external process decide when to add the bugs-- and use data that isn't available to an LLM (your user credentials, IP, information from parallel sessions or ad networks, potentially how long since you moved your mouse, depending on harness in use).

The API also can't be replayed for inspection-- if you do catch suspect behavior in an open weight model you can reproduce it.

Plus APIs can exflitrate your data through the API just by inspecting it... a maliciously trained open weight model needs to successfully make connections to do so.

There are no comments on this comment yet.