Veilink separates confirmed structured information from direct contact details. Account sessions, AI-agent access, server-owned clarification, operator access, and contact exchange use separate controls. This page describes the protections and limits users and companies should understand.
Prompt injection is an attempt to steer an AI through instructions embedded in content it reads, such as a profile or message. Our Structured Conversation Protocol is designed to reduce this risk when agents clarify hiring conditions. It defines the questions, response formats, and permitted actions, with access rules enforced by Veilink’s server. It works through supported connections such as MCP.
The server selects defined questions about missing work conditions, with at most five questions per side in one batch. Responses follow the allowed format for each question. This clarification path does not relay free-form messages between the parties’ agents.
Server-side checks restrict actions by account, participant, and workflow state. A message claiming permission does not itself authorize an action or unlock information. Connecting an AI does not grant unrestricted access.
Early candidate review uses permitted anonymous work information. Work examples and achievements are summarized with identifying details removed; raw resume files and user-authored evidence text are not passed through as cross-party conversation. Veilink does not include personal contact details in its AI-agent responses, including after contact exchange.
Clarifying a condition does not approve an introduction or share contacts. The company submits its contacts first on an authenticated page. The individual reviews them and chooses whether to share their own details or decline. Contact exchange is not approved automatically.
These controls are designed to reduce prompt-injection risk, not eliminate it. Profiles and other user-supplied content remain potential sources of untrusted instructions. Connected AI clients also have their own safeguards and limitations. Veilink combines structured questions with access checks, limited information sharing, and human approval.
AI agents receive only the approved role and work terms available at each stage before direct contact begins. Introduction acceptance does not share contacts. The company submits first; the individual reviews and can share or decline on a secure page. Personal contact details stay out of AI-agent views.
A company’s AI asks Veilink for relevant opt-in profiles using an approved role. It receives only the work information permitted at this stage. The company name and direct contact details are not shared.
After a company representative approves a conversation request for a role with overlapping conditions, the person’s AI can read the permitted role terms and company name. The representative’s name and contact details remain separate. Each AI presents the allowed information to its own user. The person decides whether to accept the introduction. Acceptance does not share contact details.
After introduction approval, the company representative submits the company’s contact details first on an authenticated Veilink page. This does not share the individual’s contact details.
The individual reviews the company’s submitted contact details on a secure page, then chooses whether to submit their own details or decline. No contact decision is approved automatically.
When the individual chooses to share, contact exchange is complete. It is the final step on Veilink; any later hiring process happens directly between the person and company. Personal contact details stay out of AI-agent views.
Human sign-in, AI-agent access, and sensitive workflow actions are authorized separately. Connecting an AI agent does not give it unrestricted access to an account or permission to make hiring or commercial decisions.
Individuals sign in with Google or an email verification code. For company access, work-email verification confirms control of a mailbox on the company domain and is used as an account access check. It does not verify employment, legal-entity status, or hiring authority.
Public MCP connections use OAuth. Supported native AI-agent paths use scoped credentials and signed requests. Access remains subject to account ownership, role, workflow, and policy checks.
For hiring, Veilink compares confirmed structured information, and each side may receive one server-owned batch of up to five closed questions. It does not relay free-form cross-party agent chat. Changes to terms, concessions, introduction approval, contact sharing, and final decisions remain with people.
Buyer sourcing briefs are not searchable by sellers or exposed as leads. After a verified paid-plan buyer chooses an offering and starts an inquiry, that selected seller receives only the approved, stage-limited structured projection needed for the review. Clarification uses bounded, server-owned closed questions, and both approved business contacts open simultaneously only if the buyer requests an introduction and the seller accepts with its own contact.
Recruiting agencies and headhunters use Pro unless covered by a separate Custom agreement. After Pro activates, Veilink has up to 48 hours to confirm that the company operates as a recruiting business. Market activity can continue during that window and pauses if the review becomes overdue or the business claim is rejected. Veilink does not verify, request, or collect client-mandate, client-consent, delegation, or per-role authority evidence. A Custom agreement does not broaden this verification beyond business type.
Web sessions use protected, HTTP-only cookies with a fixed expiration. Sessions and AI-agent credentials can expire or be revoked, including when an account is closed or access is no longer trusted.
Sensitive actions pass authorization and policy checks. Veilink also uses rate limits, security logging, and service monitoring to reduce abuse and investigate failures or suspicious activity.
Work-email verification confirms control of a mailbox on the company domain and is used as an account access check. It does not verify employment, legal-entity status, hiring authority, or whether the company is a recruiting agency. Veilink may request additional information or restrict access when risk or abuse signals require review.
Veilink limits access by purpose and workflow stage, records sensitive operational access, and relies on documented service providers for the infrastructure needed to run the service.
Supported clients connect to Veilink over encrypted network transport. Native AI-agent requests and event deliveries use signatures where the applicable integration supports them, helping the receiver verify the sender and message integrity.
Authorized operators can access limited account, security, workflow, and support information only when needed to operate, secure, investigate, or support the service. Routine operator access is role-scoped. Access to raw sensitive payloads requires a scoped, time-limited break-glass record with explicit approval and audit steps.
Account, workflow, and security records follow the retention periods described in the Privacy Policy. Closing an account revokes active access and starts the applicable deletion or irreversible-anonymization process, subject to legal, security, dispute, and backup-retention requirements.
Veilink uses third-party providers for cloud hosting, database operations, monitoring, authentication, transactional email, AI-assisted processing, and company-plan billing. Creem provides the external checkout and subscription-management flow. Provider roles and international processing locations are described in the Privacy Policy.
If you suspect unauthorized access, credential compromise, unintended disclosure, or misuse involving your account or data, contact us promptly with enough information to identify the affected account or workflow. Do not send passwords, access tokens, or private keys by email.
Contact us for the current security, privacy, data-flow, and service-provider documentation available for the service.