01Our approach
Our customers handle sensitive records: claims, cases, supply and field operations, public services. We design CaseFlow on three rules: every resource has one place of record, access follows each person’s roles and clearance, and nothing with a consequence happens without an accountable person.
02Identity and access
- Every user signs in as themselves; shared accounts are not supported.
- Roles and per-person clearance decide which objects and fields each person can see.
- Administrators can change or withdraw a person’s access at any time.
- Where a deployment serves several organizations, each organization’s data is kept separate; every hosting option runs the same code.
03People approve effects
- Sending a message, changing a record, publishing an ontology or activating an automation requires an authenticated person’s decision, recorded by the service that owns the resource.
- A model score or an agent’s explanation is evidence. It never authorizes an effect on its own.
04Agents and connectors
- The agent runtime starts empty. An organization attaches specific connector releases to a specific agent, and unattached tools are invisible to it.
- Connectors are installed as immutable releases, each pinned by a cryptographic digest. A release that does not match its approved digest or tool list fails closed.
- Secrets are injected into a connector only when it starts. They are never packaged with it, never placed in an agent’s context and never written to logs.
- Each connector runs with least privilege and bounded CPU, memory, time and output. Tools that read are distinguished from tools that change data.
- Content from connected systems is treated as untrusted input, so instructions hidden in an email or file cannot grant an agent new authority.
05Data protection
- All traffic to the Services and to this website is encrypted with TLS.
- Customer Data stays in the deployment the Customer chooses: its infrastructure, its cloud or a dedicated environment.
- Our staff hold no standing access to Customer Data.
- Customer Data and Google user data are never used to train AI models.
- Provenance, ontology version and approval are recorded with each result.
06Where it runs, and who can reach what
- On your servers or in your cloud account. You operate the infrastructure. Pannoesis has no access unless you grant it for a support request.
- In a dedicated environment we run for you. Named Pannoesis operators manage the infrastructure. They have no standing access to your data inside CaseFlow, every infrastructure action is logged, and access to your data still needs your written grant.
07Encryption and keys
- All traffic to the Services and to this website uses TLS.
- Where you operate CaseFlow, your storage encryption and key management apply, and we document what to enable.
- Where we operate it, storage encryption and who holds the keys are set out in your Order Form.
- Connector credentials are kept in the deployment’s secret store and given to a connector only when it starts.
08Backups and recovery
- Where we operate CaseFlow, backup frequency, retention and recovery targets are set out in your Order Form.
- Where you operate it, your own backup policy applies; we document what must be backed up to restore it.
- Deleting Customer Data includes its backups, on the schedule in the Privacy Policy.
09Operations and logs
- Changes are reviewed and tested before release; releases carry build evidence.
- Production deployments follow a single controlled path, separate from development tooling.
- Actions, approvals and agent tool calls are recorded in audit logs available to the Customer.
- Audit records are kept for as long as the Customer’s retention settings say.
- We maintain an incident-response process and notify affected Customers without undue delay.
10Security testing
- Every change is reviewed and tested before release.
- We have not yet commissioned an independent penetration test. When we do, we will say so here, with its date and scope.
- Reports from outside researchers are welcome under the policy in section 12.
11Third-party attestations
We have not yet published a third-party certification or audit report, such as ISO/IEC 27001 or SOC 2. We will list each attestation here, with its scope and date, once it is issued — not before. Customers can request our security questionnaire responses under a confidentiality agreement.
12Report a vulnerability
If you believe you have found a security vulnerability in our website or Services, email contact@pannoesis.com with the subject “Security report”. Please include the affected address or component, the steps to reproduce it, and its likely impact.
What we commit to
- Acknowledge your report within three business days.
- Keep you informed while we investigate and fix it.
- Credit you when it is resolved, if you wish.
- Take no legal action against research done in good faith under these rules.
What we ask of you
- Do not access, change or delete data that is not yours; stop and report as soon as you reach any.
- Do not run denial-of-service tests, social engineering or physical attacks.
- Give us reasonable time to fix the issue before disclosing it publicly.
Our contact details are also published at /.well-known/security.txt.