AI risk changes when output becomes action
The moment a system can act rather than answer, the consequences of being wrong change category entirely.
A wrong answer is a wrong answer. A wrong action sends an email, moves money, deletes a record or changes a configuration. The model has not become less reliable; what it is connected to has changed. That is the line worth governing around, and it is crossed quietly when somebody adds a tool.
More on AI and LLMs
- An LLM predicts plausible continuations, not verified truthIt only checks the shape
- Context is temporary working material, not permanent knowledgeThe board gets wiped
- Retrieval adds documents, not guaranteed correctnessThe filter slot is empty
- Temperature changes variation, not factualityThe dial only sets the spread
- System prompts are instructions, not a security boundaryA sign, with no fence
- LLM output is untrusted input downstreamIt comes in round the back