AI risk is business risk. We conduct active research into what AI systems can expose, create, or quietly get wrong— and the guidance to navigate easily.
In practice
AI risk is business risk. It shows up where models touch operations, where generated output becomes action, and where reporting hides what actually happened. We map those surfaces before they become incidents.
01
Exposure measured before it is managed.
We start by finding where models already touch the business— including the uses nobody approved. The first map is honest, so the path to reduce exposure starts from evidence.
02
Prompt surfaces, permissions, and action paths.
Risk concentrates where generated output becomes action: prompt surfaces, tool permissions, data exposure paths, and the hand-off from answer to execution. We look for the places a plausible answer can do damage.
03
Permissions and scope in the structure.
The safest limits are the ones input cannot rewrite. Permissions and scope belong in the system's structure, so a convincing prompt cannot talk an agent past them.
04
What ran, what failed, what needs a human.
Risk review starts from evidence, not logs nobody reads. What ran, what failed, and what needs a human should be visible enough to act on.
Engage
We take on a small number of engagements, so responses may be selective.
Contact usCookie preferences
Strictly necessary
Required for the site to work. Remembers your cookie choices. Always on— no tracking.
Preferences
Remembers settings between visits. Stores your light/dark/system theme choice on this device.
Analytics
Helps us understand how the site is used. First-party measurement of page views, sections viewed, and interactions (with a per-tab session id), plus Google Analytics (GA4). Nothing loads until you opt in. Your choice carries over to psysig.org.
Details on what each category stores are in our privacy policy.