The Enterprise AI Module
The exact questions enterprise procurement teams now append to standard security reviews, and what a good answer looks like.
Q: What is the provenance of the models used in the system?
Document model origins (open source, proprietary, vendor API) and versioning. Mapped to ISO 42001 Control A.8.
Q: What are the training data rights and copyright indemnification provisions?
Establish whether customer data is used to train broad models, and verify vendor IP indemnification. Mapped to ISO 42001 Control A.9.
Q: What defences exist against prompt injection and jailbreaking?
Input sanitisation, isolation of system prompts, and adversarial testing logs. NIST AI RMF Manage function.
Q: How are hallucinations and factual inaccuracies controlled?
Grounding techniques (RAG), confidence scoring, and human-in-the-loop overrides. NIST AI RMF Measure function.
Q: List all AI subprocessors involved in data processing.
Clear inventory of OpenAI, Anthropic, AWS Bedrock, etc. SOC 2 CC9.2 / ISO 27001 A.5.19.
Q: Have the models been evaluated for bias and discriminatory outputs?
Provide red-teaming reports and bias assessment metrics. NIST AI RMF Map function.
Passing this module requires evidence.
Get an exact
figure.
Fill out a short questionnaire with your company size and framework needs. We return a binding fixed-fee proposal within one business day.
Request a quote ↗