Independent consulting
AI security review for healthcare and business
Before an AI rollout handles sensitive information, establish where the data goes, who can access it, and what the system can do on someone else's instructions.
Who this is for
For healthcare and medical technology leaders, engineering teams, and business owners evaluating an AI deployment or a vendor's security claims.
Questions the work should answer
- Where do prompts, retrieved documents, outputs and logs travel?
- Can a user retrieve another person's or organization's information?
- Can an untrusted document cause the application to take an unauthorized action?
- What evidence supports the vendor's claims about retention, access and isolation?
What you receive
A data-flow map, findings ranked by severity, evidence and limitations, a prioritized remediation plan, and a leadership readout.
How the evidence supports a decision
A useful review distinguishes an observed weakness from an untested concern. Each finding should say what was examined, what the evidence shows, what remains unknown, and who can act on it. The leadership readout connects those findings to the decision about the deployment.
What to bring to the first conversation
Start with the business use case, a diagram without secrets, the systems in scope, and the decision deadline. We agree access and protected data handling before you provide records or patient information.
Start with the questions.
Use the free AI vendor worksheet to identify missing evidence before your next review conversation.
Open the AI vendor worksheetBefore you commit
Does running an LLM locally make it private?
Location alone does not establish privacy. Access, retrieval, logging, backups, connected tools, update paths and permitted data use all need examination. The review follows the actual deployment and its evidence.
Does this certify a medical AI product?
No. This is a defined security review. It does not establish clinical accuracy, regulatory clearance or legal compliance.
Work you can inspect
Read before the conversation.
View the fictional decision memorandumYour next step
Tell me what needs to happen.
Describe the problem and your deadline. I confirm fit, availability, scope, and fees before you commit.