Security and Trust

xFalcon runs as a governed layer over the warehouse you already operate. It holds a read-only connection to the databases you nominate - Snowflake, Databricks, BigQuery, Postgres, Redshift, or SQL Server - and can query only the tables your data stewards have registered.

There is no migration project and no copy of your warehouse held by Iseyon. Every answer is produced by querying your live data at the moment the question is asked, which is also why any answer can be traced back to the rows behind it.

• Your warehouse remains the system of record. Connections are read-only and row-limited.
• Iseyon holds no replica of your warehouse.
• The hosted platform runs on AWS, encrypted in transit and at rest, across multiple availability zones with cross-region disaster recovery.
• If hosted infrastructure is not acceptable, a signed installer runs the full platform inside your own network, including fully offline and air-gapped deployments.

• Single sign-on over OIDC, including Okta and Azure AD.
• Role-based access control with per-tool permissions.
• Row-level security applied to queries.
• Table visibility is set by your data stewards through a registered catalog. xFalcon cannot see a table that has not been registered.
• A full audit trail records access and queries.
• One admin console governs memories, annotations, data sources, and users, so there is a single place to review what the platform knows and who can use it.

You choose the model and you own the provider relationship. By default xFalcon uses your own key - your existing Claude, OpenAI, or Gemini account - so model calls run under an agreement you already hold and spend you already control. Iseyon-managed access to those providers is available if you would rather not manage keys.

Where no data may leave your premises, a dedicated on-premise GPU server runs open-source models instead, with no per-user AI fees.

Accuracy is treated as a control, not a claim:

• Every result is tested against live warehouse data before it is returned.
• Memory keeps corrections permanently, and annotations carry your data team's guardrails into every query.
• Self-correction checks each query before and after it runs.
• Every number drills through to the tables, rules, and rows behind it, so any answer can be re-derived by hand.

These controls reduce error. They do not remove your review obligation - see the accuracy disclaimer below.

The hosted platform is a SOC 2-designed environment: encrypted everywhere, multi-AZ, cross-region disaster recovery, and zero-downtime deploys. Database credentials are read-only and scoped to the registered catalog.

"SOC 2-designed" describes how the environment is built and operated. Ask us for the current status of formal attestations and we will tell you plainly where we are.

The information, software, documentation, analytics, and outputs provided through xFalcon may contain confidential and proprietary information. Users are responsible for ensuring that only authorized personnel access the platform and any connected or uploaded data.

By accessing and using xFalcon, users acknowledge their obligation to maintain the confidentiality of all proprietary information, datasets, credentials, documentation, and intellectual property made available through the platform. Unauthorized disclosure, reproduction, distribution, or use of confidential information is strictly prohibited and may result in legal action as outlined in applicable agreements, including any executed Non-Disclosure Agreement.

xFalcon provides AI-assisted analytics and decision-support capabilities. While validation, business rules, and quality controls are applied to improve accuracy, xFalcon does not guarantee the completeness, accuracy, or suitability of any output for a particular business, legal, financial, regulatory, or operational purpose. Users remain solely responsible for reviewing, validating, and approving all decisions derived from platform outputs before relying upon them.

To the maximum extent permitted by law, Iseyon Analytics, its affiliates, employees, contractors, and partners shall not be liable for any direct, indirect, incidental, consequential, special, or punitive damages arising from the use of, reliance upon, or inability to use xFalcon, its documentation, analytics, recommendations, or outputs.

Security questionnaires, architecture reviews, and questions about these disclosures: info@xfalcon.ai