AI / Agentic AI Interview Questions
What is Context Poisoning?
Context poisoning is a more subtle variant of prompt injection where an attacker doesn't inject an obvious command, but instead subtly manipulates a document or data source an agent will later read, altering the agent's behavior indirectly.
- Harder to detect than a blatant injected instruction, since the manipulated content can look like ordinary data
- Can influence an agent's later reasoning or tool choices without ever appearing as an explicit command
- Particularly relevant for agents that build up long-term memory from external sources over time
Because the manipulation is subtle and can accumulate silently in memory, context poisoning is one of the harder agentic security risks to catch through simple keyword-based filtering.
How does context poisoning differ from an obvious injected command?
Why is context poisoning hard to catch with simple keyword filtering?
More Related questions...
