Meet Orbital.

Orbital is Aona’s own security technology: purpose-built models, deterministic detectors and the evaluation work behind them. It powers sensitive-data recognition and policy checks inside Aona, with supported processing on employee machines, on-premises or in the cloud. Orbital 2 runs the whole check, Radar, Lens and Reflex, on a laptop in under a tenth of a second. Orbital Edge and Orbital Cloud are the Orbital Models, each built from the same four specialist Submodels: Radar, Lens, Reflex and Vision. Availability depends on the release and configuration.

Orbital’s role here is AI security: detecting sensitive data and checking requests against policies on supported employee and agent use paths. It is not a replacement for the AI tools your organisation uses to write, research or code.

Orbital runs Aona’s own model inference instead of requiring a paid general-purpose AI API call for every Orbital check. Deterministic checks need no model inference; focused models handle recognition and policy context. This architecture is designed to reduce inference spend. Actual savings depend on the workload and deployment; compute, hosting and Aona service costs still apply.

Structured identifiers need validation. Personal data needs recognition. Business policies need context. Orbital brings these different techniques into one security system. Aona evaluates both missed detections and false positives using labelled examples, hard negatives and held-out tests. For your policies, test separate allowed and disallowed requests on the intended Model, language and deployment. This is an engineering approach to accuracy, not a promise of perfect detection or a universal benchmark score.

Aona measures Orbital and the alternatives on the same held-out data. On 887 held-out policy prompts across 52 policies, Orbital 2.2.9 made 96.6% correct decisions, against 80.5% for Granite Guardian 3.1 2B and 76.3% for Llama Guard 3 1B, each given the same written policy. Hosted frontier models scored 96% to 99% but refused 20 to 30 of the prompts. On personal data, Radar and Lens raised 9 false alarms on 403 look-alike identifiers, against 96 for Microsoft Presidio and 242 for GLiNER-PII, and scored F1 0.936 on names, places and dates across 12 languages, against 0.271 and 0.605. Orbital was trained on Aona’s policy distribution with these prompts held out; the comparison models judged them zero-shot. Review the method notes, then evaluate your own workload.

No. Policy accuracy measures reference flag-or-not-flag decisions. The separate false-alarm test measured personal-data detections on 240 ordinary business documents with no personal data: Orbital 2.2.9 flagged 1 (0.4%). That is not a general hallucination rate. No matched cost-per-request comparison is available here. Orbital avoids a separate general-purpose AI API call for its own inference; compute, hosting and Aona service costs still apply.

Orbital supports deterministic rules and model-based checks. Use explicit patterns, term lists and checksum validation when the rule is precise; use learned recognition or Reflex when meaning and policy context matter. Model-based does not necessarily mean random: inference can be repeatable, but a learned verdict still needs evaluation. Configure the supported checks for your task; this is not a promise of automatic escalation whenever a model is uncertain.

Orbital runs Aona’s own model inference instead of requiring a paid general-purpose AI API call for every Orbital check. Deterministic checks need no model inference; focused models handle recognition and policy context. This architecture is designed to reduce inference spend. Actual savings depend on the workload and deployment; compute, hosting and Aona service costs still apply. Deterministic-only operation avoids model inference, and selecting supported checks lets you avoid unnecessary model work. Compare total costs, including infrastructure and service costs, on representative request volumes; there is no universal savings percentage.

Orbital can identify sensitive data and evaluate requests against configured policies. On a supported Aona client path, those findings can inform redaction or blocking. For an API integration, your application must apply the returned verdict. A model check alone does not intercept every AI request.

Yes. Aona supports customer-authored policy guardrails with policy instructions and labelled examples, answered by Reflex, plus Radar patterns and term lists. This configures the check; it is not per-customer model retraining. Use separate test requests to evaluate allowed and disallowed behaviour before activation, and confirm support in your deployment.

For the check itself, yes. With Orbital Edge, Radar, Lens, Reflex and Vision run in a local agent on the employee’s machine, so the prompt is checked where it is typed and is not sent anywhere to be evaluated. Sensitive data is redacted or the request is blocked before it leaves the device. An allowed request then goes to the AI tool the employee chose, under your policy; confirm logging and telemetry settings for your deployment.

Where the check runs. Both run the same Radar, Lens, Reflex and Vision Submodels: Orbital Edge uses the employee’s machine, Orbital Cloud uses server infrastructure such as a T4 GPU. Model selection does not imply feature parity across clients or deployments.

Aona offers separate choices for backend hosting and prompt processing. The backend can run in your cloud, on-premises or on Aona-managed servers. Prompt processing can run on the user device, in your environment or on Aona-managed servers. Confirm the supported Orbital Model, feature and integration configuration for your deployment.

Orbital Files prepares text from supported documents and OCR paths for configured guardrails. It is a separate processing pipeline, not an Orbital Model. Vision inspects original pixels for signature-shaped regions and remains staging-only, not generally available. Neither signature detection nor a policy verdict proves identity, consent or compliance. Confirm file formats, original-byte access and the processing path.

The documented release-quality language scope is English, French and Spanish. In Aona’s held-out tests Orbital 2.2.9 scored 99.1% accuracy across 2,802 English, French and Spanish rows, caught 98.8% of non-English attacks, and found names, places and dates across 12 languages at F1 0.936. Other languages are best effort and need use-case-specific evaluation. Test real language patterns, document types and policy examples from your organisation; release-quality scope is not a guarantee of perfect results.

No. Orbital’s use-case analytics describes the activity behind a request, such as research or email. It is separate from the configured guardrail verdict. The documented API integration is flag-gated staging, not generally available; confirm its availability for your deployment.