Vérifiez votre IA fournisseur, modèle, conformité, risques, sous-traitants, transparencefournisseurmodèleconformitérisquessous-traitantstransparencefournisseur

ADVANCE.AI's AdvanGuard suite provides AI-powered identity verification across four modules: Identity Verification (end-to-end onboarding), Document Verification (ID authenticity/forgery detection with data extraction), Face Authentication (facial biometric matching plus liveness/anti-spoofing detection), and Know Your Business (KYB/corporate verification). A dedicated AdvanGuard variant serves the crypto industry for global KYC compliance. Face Authentication has passed independent iBeta PAD Level 1 and Level 2 testing (ISO 30107-3) for presentation-attack detection -- the first in Southeast Asia to do so. Also powers alternative/AI-based credit scoring for digital lending across the broader Advance Intelligence Group.

No AI or machine-learning product or feature was found anywhere on Prograd's site. Prograd is a consumer rewards app that pays users cash for playing games and completing offers/surveys from an offer wall, with no algorithmic matching, personalization engine, or automated decisioning described or evidenced on the platform. This entry was added per platform policy to include all onboarded vendors in the AI catalog for completeness tracking, not because any AI use case was identified.

Stitch is a Banking-as-a-Service platform for banks, fintechs, and non-financial companies, covering cards, lending, wallets, deposits, payments, and ledgers via a single integration. The vendor markets the platform as 'AI-native' on its homepage, but no specific named AI feature, model, or capability (e.g. credit scoring, fraud detection) is described anywhere on the primary marketing site. This entry reflects the vague, unverified nature of that claim pending more specific vendor documentation.

Cranium ONE is an AI security and governance platform run as a continuous 'Discover, Observe, Govern, Secure, Prove' loop across an enterprise's AI models, agents, and vendors. Discover builds a living AI Bill of Materials (AI-BOM) and detects shadow AI; Observe monitors live model/agent behavior with 100+ risk signals and 250+ intent classifications; Govern maps and enforces policy to NIST AI RMF, the EU AI Act, and ISO 42001; Secure red-teams models/agents using MITRE ATLAS and OWASP attack libraries with auto-remediation (Arena Shield); Prove generates real-time 'Cranium AI Cards' as shareable, audit-ready compliance evidence via Trust Hubs. Cranium's own platform uses AI/automated analysis on customer-provided data under contract, and separately uses AI-assisted tools (analytics, chat, lead routing) in its own marketing/support operations -- both disclosed in its AI & Automated Decision-Making Notice.

Auto Classmate generates differentiated, standards-aligned lesson plans, units, and classroom resources using OpenAI's API, driven by a teacher-built 'class profile' (aggregate ELL/IEP/504/gifted composition, reading-level spread) rather than individual student records. 'ClassMate AI' is a context-aware chat co-teacher with multi-turn memory across a teacher's own lessons, classes, and logged reflections. Vendor explicitly designs against FERPA-covered student PII: no field collects student names, generation prompts instruct the model never to use or invent names, AI sub plans pass an automatic name-scrubbing step, and saved class-profile text undergoes an automated student-privacy review -- all stated by the vendor as 'safety nets, not guarantees,' since free-text and file uploads are not reliably screened.

Lark operates its own AI infrastructure for real-time translation, live transcription, and AI meeting minutes. Uses Microsoft Azure OpenAI for AI model inference, Microsoft Speech to Text, and Microsoft Form Recognizer. Data hosted in Singapore (Lark) or Beijing (Feishu) with separate data storage.