Context Poisoning
ELI5 — The Vibe Check
Context poisoning is when bad information makes it into the AI's context and contaminates everything after. One wrong fact from earlier in the session, and the model keeps citing it confidently forever. Clean or restart.
Real Talk
Context poisoning is a failure mode where erroneous information (wrong file contents, failed tool results, incorrect user corrections) enters the model's context and influences subsequent reasoning. Once poisoned, the model may propagate errors. Mitigations: validate tool outputs, cleanly report failures, reject context injection at boundaries, and restart sessions when poisoning is detected.
When You'll Hear This
"The agent kept referencing a deleted function — context poisoning from an old read." / "Our agent loop detects context poisoning and auto-resets."
Related Terms
Context Window
A context window is how much text an AI can 'see' at once — its working memory.
Hallucination
When an AI confidently makes something up — like citing a library that doesn't exist or generating code that calls a function that was never written.
Prompt Injection
Prompt injection is the SQL injection of the AI world.