NIST published its AI Risk Management Framework in January 2023. It’s voluntary, and its four functions make a good checklist.
Govern
- Who, by name, answers for this system’s errors?
- What can it do without a person approving it?
- Will you put data handling in writing?
Map
- Which systems and data will it touch?
- Which other companies process our data?
- Does our data train any model?
Measure
- What’s the starting number, and who agreed to it?
- Was accuracy tested on our data?
- Will the report include failures?
Manage
- What gets retested when a model changes?
- How will we hear about an incident?
- If we leave, what keeps running and who owns it?
Our answers are on the security page.
Sytsma Systems · 2-minute read