Skip to content

Track AI risk as it changes

AI Risk
Management

We build and run a risk register for your AI systems, from third party tools your staff use to models you train yourself. Each system is assessed for security, privacy, accuracy and misuse, then treated and reassessed when the model, vendor or data changes. You learn which AI risks are live and who owns them.

Map, measure, manage

AI risk moves
with every update.

The cycle follows the Map, Measure and Manage functions of the NIST AI RMF, with ISO/IEC 23894 guiding the risk method. Threats are described with MITRE ATLAS and the OWASP Top 10 for LLM Applications, so findings use terms your engineers and vendors already know.

Repeats as models, vendors or data change

Step 01 · Identify

List each AI system and its context.

A register entry per AI system with its purpose, owner, data and users recorded.

  • Third party tools, embedded vendor features and internal models included.
  • Intended use and foreseeable misuse written down.

Step 02 · Assess

Rate the risks each system actually carries.

Rated risks for each system covering security, privacy, accuracy, bias and misuse.

  • Prompt injection, data leakage and excessive agency checked against the design.
  • Vendor terms on training data and retention reviewed.
  • Impact assessed for the people the output affects.

Step 03 · Treat

Decide on controls, limits or acceptance.

A treatment decision for every serious risk, with an owner and a status.

  • Controls such as human review, output limits and access restrictions chosen.
  • Risk acceptance recorded by the accountable owner.
  • Risks that need proving referred to AI Penetration Testing.

Step 04 · Review

Reassess when something changes.

A dated record of each reassessment and what triggered it.

  • Triggers set for model updates, new data, vendor changes and incidents.
  • Treatment status checked against commitments.

Step 05 · Report

Show leadership the AI risk picture.

A risk summary for leadership and the board, ranked by exposure.

  • Top AI risks and treatment progress reported.
  • New and retired AI systems noted.

List each AI system and its context.

A register entry per AI system with its purpose, owner, data and users recorded.

  • Third party tools, embedded vendor features and internal models included.
  • Intended use and foreseeable misuse written down.
Repeats as models, vendors or data change

Key benefits

Know which AI risks are live, and who owns them.

One register for all AI

Vendor features, staff tools and in house models sit in one register with consistent ratings. Leadership can compare AI risk with every other risk instead of guessing.

Risks named in plain terms

Threats are described with MITRE ATLAS and the OWASP Top 10 for LLM Applications, then translated for the business. Engineers and executives read the same entry and understand it.

Reassessed when things change

Models update, vendors rewrite their terms and new data flows in. Each of those changes triggers a reassessment, so the register describes the systems you run now.

What you get

What lands on
your desk.

What the program produces, and keeps current, through each cycle.

The document

The AI risk method

Rating criteria, assessment steps, reassessment triggers and roles, mapped to the NIST AI RMF and ISO/IEC 23894 so the method holds up under outside review.

The working file

The AI risk register

Each AI system with its risks, ratings, treatments, owners, status and last review date. The record your governance group, your auditors and your engineers all work from.

The conversation

The AI risk review

A recurring session with your AI system owners and sponsor to go through new systems, changed risks and treatment progress, with decisions recorded in the register.

Client case studies

Real engagements.
Documented outcomes.

YHB | CPAs & Consultants testimonial

In their words

"ForgePath has become a trusted security partner for YHB. Their team understands how a CPA firm operates and has helped us build a security program that’s both practical and sustainable."

Cyrus Kapadia
Chief Technology & Information Officer, YHB | CPAs & Consultants

Trusted by industry leaders.

AI risk questions

Before the
first entry.

Straight answers from the practitioners who do the work. If yours is not here, ask us directly.

Yes. Vendor AI still processes your data and shapes decisions. The risk sits in what you send it, what the vendor keeps, and what your staff do with the output. For many organizations, the register starts with tools they buy.

No. Risk management decides which risks matter and how to treat them. AI Penetration Testing, under Assure, proves whether specific attacks work. When an assessment points to a risk worth proving, we recommend a test.

An owner for each AI system, design documents and vendor terms, and time with the engineers or administrators who run each system. A sponsor on the leadership side approves treatments and risk acceptance.

AI risks can live in your enterprise register or in a dedicated register that feeds it. We use your rating scales where possible, so AI risk is weighed on the same terms as everything else.

Ready?

Bring us
the problem.

One conversation with a senior practitioner who has carried this responsibility before. We will tell you what we would do, and whether it is us.