Frequently asked questions

Plain-language answers about your data, access and first AI tools.

What data does Briard ingest?

Briard receives the information your team enters and the limited information your organization authorizes a supported connection to provide.

Information	Examples
Governance records and supporting material	AI-tool names, providers, owners, purposes, information categories, answers, policies, review decisions, approved evidence files and references
Approved discovery metadata	Supported application or resource references, bounded activity signals, source health and timestamps; permitted device and model inventory from enabled checks
Account and service information	Account and workspace membership details, policy-recipient information you supply, subscription records, delivery information and support messages

Deliberately uploaded evidence is actual file content. Briard stores permitted evidence privately and reads its bytes for integrity and file-safety checks. The configured malware-scanning provider receives those bytes for scanning. A purpose brief, configuration record, human-review procedure, completed test report or dated decision can support an inventory answer when your organization approves it for this platform. See file-evidence instructions.

Normal discovery does not collect raw prompts, model replies or business-file contents. Do not put credentials, CUI, PHI, payment-card data or customer production records in ordinary fields or evidence uploads. Connection credentials follow a separate protected setup path; card details go directly to the payment provider. Metadata can still identify people and should be treated with care.

Read Shadow AI Discovery, Fleet inventory and coverage and the data flows and responsibilities for the applicable source and service boundaries.

How does Briard keep my data safe?

The supported controls include encrypted connections, workspace and role checks, database tenant isolation, MFA for owner and administrator access, private evidence storage, file integrity and safety checks, and audit history. Connection credentials are protected on the server. A file passing its scan does not establish that its statements are accurate or that the AI tool is approved.

Briard Alert & Response Console (BARC) binds access to the authenticated tenant and authorized roles. Cases, searches, evidence and exports stay within that scope. Custody records preserve evidence relationships, digests and available signatures so a reviewer can check integrity; they do not prove a source was truthful or had complete visibility.

Customer-managed sources and investigation storage also require the approved access, network, encryption, region and retention configuration. Normal monitoring and separately authorized content investigation have different controls. The Trust Center explains the security posture and assurance evidence; BARC security and data boundaries explains the deployment and evidence controls.

What data does Briard read to provide results?

Briard uses saved tool facts, mapped answers, evidence status and references, policies, review decisions and approved discovery metadata to organize the inventory, identify documentation gaps and prepare governance records. Your team confirms proposed findings and makes the review decisions. Briard also reads uploaded file bytes for integrity and safety checks, including its configured scanning service; uploading a file is not automatic approval of its contents.

Briard's normal investigation monitoring analyzes approved event metadata: timestamps, actor and tool references, supported model or deployment references, operation types, outcomes, policy decisions, bounded counts, source health and receipts. It uses those signals to correlate activity, explain alerts and cite the evidence behind findings. References are pseudonymized where required. Raw prompts, model replies and file contents are excluded from normal monitoring.

A finding depends on the connected source's actual visibility. Missing events remain unknown; metadata alone cannot reveal what a prompt said or establish a person's intent. See Connect sources and Alerts and real-time response.

If Briard finds a true-positive alert, can its access be elevated beyond metadata for an investigation?

Yes, through separately approved access for a linked case, where the configured source and investigation setup support it. Confirming an alert does not automatically expand permissions or authorize reading content.

The Request deeper investigation workflow offers two choices:

Choice	What it permits within the approved scope
Check content and show findings	Examine approved content inside your environment and keep findings and receipts, without keeping original content as case evidence
Read original evidence	Hold only the selected originals for review. In the same case, an authorized reviewer chooses items and an access period. Briard prepares them automatically after the required independent approvals.

The request identifies the case, source, purpose, material, time window, item and size limits, expiry, and required storage and retention. Two distinct eligible reviewers other than the requester must approve. The source, privacy policy and required storage controls must also be configured and ready.

Sending a request does not start collection. Approval alone does not prove that collection ran; check the actual state, receipts and handling outcome. Users do not move files or enter storage keys. Normal monitoring remains metadata-only. Access expiry, protection from deletion, and controlled disposal are separate; legal holds may preserve evidence longer without extending permission to open it.

See Choose what BARC can access for the request, approval and recovery steps.

Why should I use Briard if my organization does not use much AI?

Start with the few AI tools you already use. Even one assistant can handle confidential information or influence an important decision. Briard gives the team one shared record of the tool, its owner, allowed use, information categories, safeguards and supporting proof.

That record helps you answer customer questions, review a vendor, onboard staff and preserve decisions when responsibilities change. You can begin with one tool and add records as your use grows. Approved discovery may also identify AI features within supported work sources that your inventory has missed; it does not establish complete coverage.

Start with Add an AI tool, complete its purpose and information categories, and attach the proof that supports the answers. The goal is a clear, current record your team can use.