Exposure Intelligence Lab · Working model v0.1
The AI Exposure Index
A practical framework for estimating how easy it may be to impersonate, exploit, or misrepresent a person or organization using publicly available information, media, relationships, and trust signals.
The thesis
The more raw material you publish, and the more people act on your word without checking, the cheaper you are to fake and the more damage a convincing fake can do. Exposure is not the same as risk. But it sets the ceiling on it. The Index is a way to think about that ceiling before someone else does.
The working model
Score each factor from 0 (none) to 3 (high), then add them up. There are thirteen factors across three groups, so scores run from 0 to 39. The number is not the point. The exercise is: it forces you to look at your exposure the way someone trying to exploit it would.
1
Synthesis surface
How much raw material already exists to fabricate you.
Public photos
Clear images of your face, from many angles, in good light.
0–3
Public video
Footage of you moving and speaking on camera.
0–3
Public audio
Recordings of your voice, especially long and clean ones.
0–3
Speaking appearances
Talks, panels, and presentations captured and posted.
0–3
Podcast appearances
Long-form audio of you in natural, unscripted conversation.
0–3
2
Reach and authority
How far a convincing fake can travel and how much authority it borrows.
Executive visibility
A title and profile that make your instructions carry weight.
0–3
Brand recognition
A name or company people recognise and trust on sight.
0–3
Search footprint
A large, easily harvested trail of public information about you.
0–3
Domain authority
Owned channels strong enough that mimicking them is worthwhile.
0–3
Public contact information
Direct lines that let an impersonation reach the right people.
0–3
3
Trust dependency
How much damage a convincing fake could actually do.
Social graph visibility
A public map of who you know and who would act on your word.
0–3
Trust dependency
How readily others move money or make decisions on your say-so.
0–3
Verification weakness
The absence of agreed rituals to confirm a request is really from you.
0–3
Reading the score
The bands below are deliberate approximations, not validated thresholds. They are there to turn a number into a conversation.
Low Exposure
0 to 9
Little public media, limited reach, and few people who would act on your word without checking. A convincing fake is harder to build and easier to catch.
Moderate Exposure
10 to 19
A real but contained public footprint. Worth adopting a basic set of verification habits before you need them.
Elevated Exposure
20 to 28
Substantial public media and reach, and people who act on your communications. Impersonation is plausible and consequential. Verification rituals start to matter a lot.
Critical Exposure
29 to 39
Abundant public voice, video, and likeness, high recognition, and high trust dependency. A convincing fake is cheap to build and expensive to absorb. Treat verification as core infrastructure.
Read this first
This is a practical heuristic for thinking about exposure. It is not a security audit, a scientific instrument, or a validated risk score. It is a working model, version 0.1, and it will change. Use it to start a conversation about exposure, not to conclude one. If you have a real, active threat, talk to a security professional, not a framework on a website.
Practical implications
A high score is not a reason to disappear. For most people with elevated exposure, visibility is the job. The move is to make yourself hard to exploit even while you are easy to imitate:
- Verification that does not rely on appearance. Agreed callback procedures for anything involving money, credentials, or access. "It looked and sounded like them" should never be sufficient on its own.
- A canonical channel. One known, hard-to-spoof place that sensitive requests are supposed to come through, so anything outside it is suspect by default.
- A culture where verifying is competence. If asking a senior person to confirm a request is treated as insubordination, you have built the exact gap an impersonation needs.
- Less unnecessary raw material. Not silence, just intent. Notice the difference between media that does a job and media that only adds to the synthesis surface.
Current questions
The parts I have not resolved, and would value being argued with on:
- How should the factors be weighted? Trust dependency probably matters more than raw photo count, but by how much?
- Does organizational exposure compound individual exposure, or partly offset it through more mature verification?
- Which verification rituals actually hold up under pressure, and which only feel safe?
- Is there a cleaner way to express the output than a single number?
Related field notes
Have a sharper version of this model? I am refining it in the open. Tell me what I am missing.
Back to the Lab →