Protect the AI recipe recommender
Threat-model a manufacturing AI system before connecting it to production decisions.
OT-adjacent AI The recommender will not write directly to the PLC, but it will influence MES recipe changes. That is enough to threat-model it. Bad input, bad identity, or bad authority can become bad product. CSF plus STRIDE Identify the flow, then ask what can go wrong Use NIST CSF to manage cybersecurity lifecycle and STRIDE to inspect the actual system design. AI does not remove old security questions; it adds new places to ask them. Shortcut It is only advice, so security can wait. The model is treated as part of the production risk surface. Advice that changes production…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in