Start by finding out
where your AI is.
We'll help you prove it.
Connect decisions, controls, and human oversight with verifiable evidence from documents, code, and models. V-PROOF turns AI governance into evidence that is preserved and can be verified by a third party.
For general management, technology, data, and legal affairs.
- Board & executive managementAccountability
- AI CommitteeDecides
- System ownersOperate and attest
-
Work has context.
A draft from V-PROOF AI is linked to a project and a document.
-
Control takes center stage.
The enabled filter checks the message before it is sent. The response is audited.
-
One person responds.
The person in charge reviews the work and records their approval.
-
What matters is sealed.
When sealing is confirmed, the file's cryptographic fingerprint is recorded as evidence.
-
The evidence can be verified.
View the record or verify a file's fingerprint. No account required.
Illustrative sequence. It does not show live activity or perform operations.
December 2. Requirements for high-risk systems listed in Annex III.
The question is no longer whether your company uses AI. It is whether you can demonstrate how you control it.
When an inspection, audit, or complaint arises, a signed policy alone will not suffice. You will need to demonstrate which system was chosen, who oversaw it, and what criteria were used. That proof is built in advance, not on the day someone asks for it.
Penalty for high-risk noncompliance: up to 15 M€ or 3% of global revenue. Source: European Commission.
Much of what is now called AI governance is still just a statement of intent.
- There is no complete inventory.
AI tools are being used that no one has set up. Without an inventory, there is no classification, and without classification, there are no controls.
- The policy lives in a document.
It says what must not be done, but nothing prevents it when the moment comes.
- Human oversight is taken for granted.
Someone checks, but there is no record of who did it, when, or what was verified. The EU AI Act requires this to be demonstrable (Art. 14).
- The record is kept by the party that has to justify itself.
An internal record, however immutable, still requires trusting whoever maintains it.
Evidence is more reliable when it can be verified without relying solely on the custodian of the record.
V-PROOF lets anyone check the recorded evidence without an account or access to the client's internal systems.
How is it verified?Governance defines what should happen. V-PROOF helps prove what did happen.
What was evaluated, what was documented, and when.
Which document, image, or video; with which fingerprint; and on what date.
Which commit, which checks, and in what context.
Which version, data, parameters, and results.
What risk am I taking, and who is liable if something goes wrong?
A documented overview of AI, including risks, responsible parties, approvals, and associated evidence. Information for the board, the audit team, and the client.
How much AI do we actually have, and how can I control it without holding anyone back?
Inventory and controls for V-PROOF workflows, code auditing with GitHub, and model auditing with MLflow. APIs and SDKs to connect your processes.
What data is being fed into external models?
The local filter checks chat messages before they are sent, when enabled. When a document is sealed, only its fingerprint is recorded: the file is not stored on the platform. During verification, the fingerprint is calculated in the browser, and the file is not uploaded.
Do I have evidence that would stand up to an audit?
Approvals with a trusted timestamp, change history, and evidence linked to controls. A third party can check the seal or verify a file without an account.
The same evidence for four worlds that are currently separate.
Risks · controls · owners · approvals
Identity · Authority · Action · Result
Repositories · commits · dependencies · checks
MLflow · versions · data · metrics · promotion
sealed document, commit, or model with its context
Fingerprint and context linked to the evidence
A third party verifies it, without an account or access to your systems
Your runtime determines what an agent can do. V-PROOF provides evidence of what it did. We don't replace your GRC, security or MLOps tools; we connect them with evidence.
See.
Control.
Prove.
AI governance, assisted writing, code and model auditing, and V-Seal, all on a single platform. Link your team’s work to your risks, owners, approvals and evidence.
Record systems, agents, models, and third-party services, with their purpose, owner and risk level. Synchronize vendors and metrics from V-PROOF and recalculate risk on demand.
76 initial controls, assessments and approvals with supporting documentation. Each owner records the status of their controls and attaches evidence.
Connect GitHub: functional glossary, history, version comparison, and knowledge and architecture maps. Checks generate evidence without blocking deployments.
Draft from the portal, Word or the desktop app. Each chat is associated with a project, document, and version, along with its corresponding conversation and audit trail.
With the filter enabled, a local classifier checks messages before they are sent to the provider and blocks them if they fail. Each response is checked against the reference corpus.
Connect MLflow in read-only mode. Run six checks on your models and detect unapproved versions in production. V-PROOF records the results; it does not modify MLflow.
Sign documents and commit or model manifests: recorded cryptographic fingerprint, encrypted metadata on IPFS, and declared provenance.
No account required. When verifying a file, its fingerprint is calculated in the browser without uploading it. The QR code checks the registry; it does not validate the file on its own.
Link evidence to systems, controls, risks, and approvals. See who did what and generate records and reports for auditors and management.
Current scope. Synchronization, risk recalculation, and MLOps controls are executed on demand. Chat safeguards apply to flows within V-PROOF. AI Gateway, MCP, and Shadow AI are part of the roadmap.
Use AI to draft the text, or register the system, document, commit, or model you want to govern.
Apply chat safeguards and code reviews or MLOps checks. Document the corresponding human evaluation and approval.
Confirm that the file or manifest has been sealed. Its fingerprint is recorded along with the reference to its encrypted metadata.
The QR code checks the record. To verify a file, calculate its fingerprint in the browser without uploading it to the platform.
Governing is necessary. Being able to demonstrate it is what makes the difference.
V-PROOF connects daily work with your governance record: assisted writing, code, models, risks and approvals. Evidence is linked to the context that explains it and can be sealed for verification.
We do not replace your GRC platform or your cybersecurity team. The portal, Word, and desktop share data and processes; the API, SDK, and Node-RED allow you to build integrations tailored to each deployment.
From an unregistered assistant to a system with risks, approvals, and documentation.
Here is an example: register a system, assess its risk, document the human review, and link evidence to its V-Seal.
View the guided case →V-PROOF in Action. Traceability applied to real-world processes.
FFUUSS / OTS Group and ROCADA have implemented V-PROOF to organize technical evidence regarding versions, revisions, approvals, and human intervention in AI-assisted design processes.
External Validation
-
“The EU AI Act shifts the risk directly to the Board of Directors.”
-
“Governance will drive AI adoption.”
-
Ensuring compliance with the European AI Regulation (EU AI Act).
-
“Enterprise AI needs a black box, not an ethics manual.”
We hold ourselves to the same high standards that we expect from our clients.
With OCA Global, an ENAC-accredited organization. See the Trust Center for details.
Start by knowing where your AI is.
We’ll help you prove how it’s governed.
What AI does your organization use, what is its risk level, and what evidence is it missing before December 2027?