What we do with your code, stated per deployment.
Written for a security reviewer. It states the training position, the data handling per deployment mode, the provider record and the certifications we do not hold.
Data handling, per deployment mode.
ManagedPlatform data is processed in the environment LatentForce operates. Inference leaves only where you permit an approved third-party model.
On-premisesPlatform data and repository content stay in your environment. Permitted external inference is the only egress, under that provider’s terms.
Air-gappedNo external model calls occur in this configuration. Data and inference stay inside the isolated environment.
Who can see what.
Team accessAccess and model tiers are configured per team. Team permissions are distinct from task routing.
Repository accessRepository context respects the access rules of the code host it was indexed from.
Human reviewGuidance and skill updates are proposals. A named reviewer approves them before they apply.
Certification status, stated truthfully.
Certifications heldLatentForce does not currently publish a SOC 2 or ISO/IEC 42001 certification. Where a certification is not held, this page says so rather than implying it.
Security questionnaireCompleted on request for an evaluation in progress, against the configuration under discussion.
Contractual commitmentsData handling commitments are in the Terms of service and the Privacy policy, not on this page alone.
A provider record per deployment.
ApproachNo single fixed subprocessor list describes every managed, on-premises and air-gapped configuration. The provider record is produced per deployment.
Model providersAny approved third-party model you permit is a provider for that workload, under that provider’s terms.
Audit trailEach request through the gateway is priced, logged and attributed by user, project, model and provider. Requests that do not pass through the gateway are not recorded there.