Skip to content
AethosHub
ProductsAgent Access ManagerOn-prem AI gateway, virtual keys, routing, guardrails, auditAPIKeyOpsAI API key inventory, lifecycle, ownership, spend evidence
Use casesPrevent AI data leaksLayer controls across prompts, responses, and tool-call argumentsControl AI spendCombine budgets, rate limits, and qualified model failoverSecure agent & app accessScoped, revocable virtual keys instead of shared provider keysDetect & contain abusePrepare, contain, investigate, and recover from a leaked key
Platform capabilitiesAI gateway & routingOne OpenAI-compatible endpoint with load balancing and failoverAudit & reportingSearchable call records and scheduled operational reportsEnterprise SSO & RBACKeycloak or your OIDC provider, role-based administrationTrust & security architectureDeployment model, data handling, and vulnerability disclosure
ResourcesProduct documentationCurrent installation, concepts, operations, and recipesSecurityArchitecture, data handling, and vulnerability disclosureImplementationRemote onboarding, onsite services, and handoverSupport commitmentsCoverage, severity levels, and response targetsCompany informationLegal identity, operating model, and contact details
BlogSupport
PricingRequest Demo
PricingRequest Demo

Trust center

Security architecture and disclosure

Last reviewed: 20 July 2026

AethosHub products are designed to keep the control plane, credentials, policies, and audit evidence inside infrastructure the customer controls. This page describes the product security model and how to report a potential vulnerability responsibly.
ProductAethosHub
VersionCurrent
UpdatedLast reviewed: 20 July 2026
ContentsSecurity principlesArchitecture & boundariesData handlingCredential protectionAccess & auditSecure operationsAssurance & documentationVulnerability disclosure

Security principles

  • Customer control: customers choose the deployment environment, network boundaries, identity integrations, providers, retention, and administrative access.
  • Credential isolation: agents and applications receive scoped virtual credentials rather than reusable upstream provider credentials.
  • Enforcement before access: authentication, routing, budgets, limits, and configured content controls are evaluated before a governed model request reaches its provider.
  • Evidence by design: security-relevant administrative and runtime activity produces customer-controlled audit records.
  • No required telemetry: normal product operation does not depend on sending gateway traffic or usage telemetry to AethosHub.

Architecture and trust boundaries

Agent Access Manager is deployed in a customer-managed environment using the supported deployment model for the licensed edition. Applications and agents connect to the customer's gateway endpoint. The gateway evaluates configured identity, virtual-key, provider, model, budget, rate-limit, and guardrail policy, then forwards the governed model request to a provider destination the customer configured.

The customer remains responsible for its cloud or datacenter controls, host and cluster security, network segmentation, TLS termination, backing services, identity provider, backups, monitoring, and the security of configured model providers. AethosHub is responsible for remediating supported-product vulnerabilities under the applicable subscription and support terms.

Data handling

Product data

Configuration, virtual-key records, policy, usage records, and audit trails are stored in the customer's deployment. Model inputs and outputs pass through the customer-operated gateway and are sent only to destinations configured by the customer. AethosHub does not receive that traffic during ordinary product operation.

Support access

AethosHub does not require standing access to a customer environment. If a customer asks us to review logs, configuration, or an environment during support or professional services, access must be expressly authorized, time-bounded where practicable, and limited to the material needed for the task. Customers should remove secrets and unrelated personal data before sharing diagnostic material. Processing of personal data during requested services is governed by our Data Processing Addendum where applicable.

Website data

Website enquiry and support-form data is handled as described in the Privacy Notice. It is separate from data processed inside customer deployments.

Credential protection

  • Upstream provider credentials are stored encrypted and are accessed by the gateway only to execute authorized requests.
  • Applications use revocable virtual keys with customer-defined identity, provider, model, budget, rate, and policy scope.
  • Model-generated tool-call arguments can be screened by configured response guardrails, but tool execution and tool credentials remain the responsibility of the agent runtime or tool service.
  • Credential rotation, expiry, storage keys, and backup protection remain under customer administrative control.
  • Sensitive values should never be placed in support tickets, screenshots, source repositories, or unencrypted diagnostic exports.

Administrative access and audit

Customers should integrate supported enterprise identity, require strong authentication, separate administrative and workload identities, assign least-privilege roles, and review privileged access regularly. Audit records should be retained in accordance with the customer's investigation, regulatory, and storage requirements and exported to its monitoring or SIEM tooling where appropriate.

Secure operations

  • Deploy only supported releases and apply security updates within a risk-appropriate maintenance window.
  • Restrict management endpoints to trusted administrative networks and expose only required gateway endpoints.
  • Use customer-managed TLS certificates, secure secret stores, encrypted backups, and tested restoration procedures.
  • Monitor authentication failures, policy denials, unusual usage, administrative changes, and destination errors.
  • Validate updates in a non-production environment and maintain a rollback plan for production changes.

See the installation and administration documentation for the deployment checklist.

Assurance and security documentation

We do not claim certifications on this page that have not been independently completed. Under an appropriate confidentiality agreement, qualified customers may request architecture documentation, a product-security questionnaire response, relevant test summaries, data-flow review, and remediation status for issues material to their planned deployment. Contact us through Get a quote.

Coordinated vulnerability disclosure

If you believe you found a vulnerability in an AethosHub website or product, email support@aethoshub.com with the subject Security Disclosure. Include the affected product and version or URL, technical impact, reproducible steps, proof-of-concept material that does not expose third-party data, and a secure way to contact you. Do not include live credentials in the initial report.

Research guidelines

  • Test only systems and data you own or are expressly authorized to test.
  • Do not access, alter, retain, or disclose another person's data.
  • Do not use denial-of-service, destructive testing, social engineering, spam, or physical attacks.
  • Stop testing and report immediately if you encounter sensitive data or create service instability.
  • Allow a reasonable remediation period before any public disclosure and coordinate timing with us.

We aim to acknowledge a credible report within two business days and provide a triage update within five business days. These are operational targets, not contractual service levels. This policy does not establish a bug bounty or promise payment. Good-faith research consistent with these guidelines will not be pursued by AethosHub merely because it identified a security weakness; this does not authorize conduct prohibited by law or by a third party.

AethosHub

AethosHub builds enterprise AI control infrastructure across API key governance, provider routing, virtual access, audit, and guardrails.

CompanyCompany informationResourcesBlogSecurityPricingGet a quoteRequest a demo
ProductsAgent Access ManagerAPIKeyOps30-day evaluation
ServicesImplementationSupport requestSupport SLAProduct documentation
LegalPrivacy NoticeEmail preferencesData Processing AddendumWebsite TermsPaid Software AgreementEvaluation Agreement
Social
© 2026 AethosHub Pvt LtdPrivacy · Email preferences · Terms