AI Vendor Risk Assessment demo
Each vendor's AI stack built from its own documents, with every claim checked against them. 1:27
What the demo shows
Assesses how vendors use AI from questionnaires, contracts and model documents
A questionnaire agent sends and reads an AI questionnaire; a document reader reads the contract, terms, DPA, model cards, trust center, SOC 2 and bias audit; a stack builder assembles each vendor's AI stack, and every layer carries what the vendor said and what its documents say. A contradiction checker marks mismatches and a verifier re-reads every source on its own; disagreements go to a person. A tiering agent proposes the tier under the EU AI Act, US law and your policy, a drafting agent turns each finding into a clause, and a change monitor reopens the stack when a vendor's terms or features change. The AI risk lead talks to the agents, checks the cited passages, and approves the tier and the changes.
Transcript
- AI Vendor Risk Assessment shows risk teams how each vendor uses AI.
- 72% of firms only partly know which vendors use AI.
- For Nadia, the AI risk lead, agents read the hiring vendor Talent Loop's documents and built its AI stack.
- Talent Loop said customer data never trains its models.
- Its terms allow it.
- Its applicant ranking also sends resumes to a third-party model it never mentioned.
- A second agent re-read every source and agreed.
- Nadia confirms. She asks why it's high.
- Hiring is high risk under the EUAI Act.
- She approves the high tier with a reason and signs.
- Each mismatch becomes a clause.
- No training on our data.
- Notice of model changes.
- Every statement is cited.
- Every decision signed. Every step logged.
- The result? Two days.
- not three weeks, and the hidden model caught before renewal.
- Built on Botminds: agents, data, screens, and governance on one platform.
Transcribed from the film’s narration. Captions are on the player.