AI Governance and AI Security

Govern how your organization uses AI, and make sure your defenses are ready for it.

Already have an IT provider? Here's how we work together.

A professional firm whose staff adopted AI tools on their own

The tools nobody approved

The challenge. Staff at a growing firm use three AI writing and note-taking tools, two on personal accounts. One summarizes client calls. The firm’s client agreements promise confidentiality, and nobody can say where those summaries are stored or whether the vendor trains on them.

The insight. Banning AI rarely works; people keep using it where nobody can see. The firm needs to know what is in use, decide what data each tool may see, and check vendors the way it already checks any other supplier of client data.

The decision. The partners chose to approve a short list of tools under firm accounts, with a written rule on client data, rather than a blanket ban.

The action. We built an inventory of the AI in use and classified each tool by the data it touches. We reviewed each vendor’s terms on retention and training, wrote a one-page AI use policy with the partners, and set the IT provider’s requirements: firm accounts only, single sign-on, and logging. We also checked that the firm’s security tools could see the new AI traffic.

The result. Two tools were approved, one was replaced, and every AI tool now runs under a firm account. The partners can answer a client’s question about AI in one page.

What Security Medic does

Security for AI

An AI system inventory and classification, a data privacy and security risk assessment, AI governance policies, and due diligence on vendor AI products.

AI for security

Evaluation and oversight of AI-assisted security tools, an assessment of AI-aware threats, and security operations ready for AI-driven attacks such as convincing fraud emails and cloned voices.

Independent advice

We do not sell AI security products or managed AI services, so the architecture advice is independent of any vendor.

Privacy built in

What personal data an AI tool may see, how long it is kept, and how notices and rights requests cover it.

Where this sits in the cycle

Every client goes through the same six steps, from a town office to a family at home. The Security Medic method

  1. 1. Govern
  2. 2. Design
  3. 3. Assess
  4. 4. Review
  5. 5. Conform
  6. 6. Improve
  • Govern. Decide who approves AI tools and what data they may use.
  • Design. Set the target: approved tools, accounts, logging and data rules.
  • Assess. Inventory what is in use and measure it against the target.
  • Review. Rank the gaps by the data at risk, and decide with the leaders.
  • Conform. The IT provider or partners make the changes; we check them.
  • Improve. Review new tools and new threats each quarter, and raise the bar.

The standards we draw on

The NIST AI Risk Management Framework and its Generative AI Profile (NIST AI 600-1), the NIST Privacy Framework, the NIST Cybersecurity Framework 2.0, and the OWASP Top 10 for LLM Applications for AI that the client builds.

Threat modelling →Third-party risk →Policy and SOP writing →

Getting ahead of it

New AI features arrive inside tools you already own. Add an AI check to the steps for approving any new software or renewal.

Start with a free 30-minute call

Tell us which AI tools your people use. We will talk through what to govern first.

Book a free 30-minute call