AI assists the investigation
PacketSafari treats AI as an investigation component, not the source of packet truth. AI-assisted features are governed through feature review, provider assessment, explicit routing, user-facing disclosure, and operational guardrails. Deterministic packet tooling establishes the evidence behind material conclusions.
The controls below show how model access, packet evidence, human review, result labelling, and private deployment boundaries are handled.
Controls customers can rely on
| Control area | PacketSafari control | Status |
|---|---|---|
| Feature purpose | AI is used to assist packet investigation and reporting, not to replace packet evidence | Implemented |
| Evidence grounding | Material findings link back to deterministic packet facts such as frames, filters, flows, timestamps, and decoded fields | Implemented |
| Human review | Reports retain evidence anchors, uncertainty, and investigation history so an analyst can challenge a result | Implemented |
| Result lifecycle | Preliminary findings are labelled separately from later verification outcomes | Implemented |
| Provider routing | AI traffic follows the provider configured for the enabled feature and deployment | Configured by deployment |
| Data minimization | The model workflow receives selected investigation context rather than unrestricted access to the capture store | Implemented |
| Private deployment | On-premises customers can control network egress and select an OpenAI-compatible endpoint | Customer controlled |
Regulatory scope
PacketSafari has not received an AI certification and does not assign one regulatory classification to every customer deployment. Classification depends on the feature, intended use, operator, and deployment. Private-model compatibility, output quality, latency, and capacity are validated for the customer-selected endpoint and environment.