Frequently Asked Questions.

1. How does Revir govern the use of AI in Scout?

Revir treats AI as a governed capability within Scout rather than a collection of standalone features. AI models and services are evaluated before being made available based on factors such as security, deployment availability, performance, accuracy, cost and suitability for justice and national-security workflows.

Revir’s approach is designed to allow AI capabilities to evolve while preserving customer control, security boundaries and established workflows.

2. Where is Scout’s AI processing performed?

Scout performs AI processing within the customer’s controlled cloud architecture, so agencies have transparency over their own security, networking, encryption and access policies to the systems processing their files and data.

3. Does customer data leave the agency-controlled environment for AI processing?

Scout is designed to favor AI capabilities that can operate within the customer-controlled security boundary, including models available directly within AWS or deployable into the customer environment on Kubernetes. Revir documents the architecture and data path for AI services so customers can understand where processing occurs before enabling a capability.

4. Does Revir use customer files, prompts or AI outputs to train AI models?

Revir does not use customer files, data, or prompts to train models unless a customer explicitly authorizes such use.

Where Scout uses third-party AI services or models, opt outs are applied to the account, or are native (i.e. GovCloud). Revir’s deployment approach is intended to give agencies visibility into and control over those choices.

5. How does Scout protect sensitive justice data during AI processing?

Scout applies the same security principles to AI processing that it applies to the rest of the platform.

Customer environments are isolated, data is encrypted in transit and at rest, customer-managed encryption keys can be used, and access is governed through the customer’s identity and security architecture. AI services are evaluated not only for capability, but also for where data is processed and how that data is protected.

6. How are AI models evaluated before they are made available in Scout?

Revir evaluates models against the workloads they are intended to support rather than assuming one model is best for every task.

Evaluation may include accuracy, latency, security, deployment availability, supported file types, context limits, cost and performance on justice-specific media and documents. This allows Scout to use different models for different workloads while presenting those capabilities through a consistent platform experience.

7. Can agencies control which AI models or services are used?

Scout’s architecture is designed to support multiple AI models and services rather than permanently tying an agency to a single provider.

8. Is AI built directly into Scout workflows, or do users have to export files to separate AI tools?

AI is integrated into Scout’s evidence and intelligence workflows.

Users can apply capabilities such as transcription, translation, OCR, computer vision, and natural-language discovery to information already managed in Scout rather than manually downloading sensitive files, uploading them into separate consumer AI tools and then moving the results back into the case or investigation. Scout eliminates these kinds of motions which are both inefficient and insecure on many levels.

9. What types of files and data can Scout analyze with AI?

Scout can apply AI capabilities across unstructured information including video, images, audio and documents.

Depending on the file and configured capabilities, Scout can generate or extract information such as transcripts, translations, OCR text, objects, actions, scenes, locations, metadata and AI-generated summaries. These results can then become searchable alongside the original files.

10. Does Scout support natural-language search across evidence?

Yes.

Scout is designed to help users search both files and the information extracted or generated from them. Natural-language discovery can help users find relevant content across transcripts, documents, metadata and AI-derived information without requiring them to know the exact filename, folder location or keyword used when the evidence was originally collected.

11. How does Scout support human review of AI-generated results?

Scout uses AI to assist people in finding, organizing and understanding information; it is not intended to replace professional judgment.

AI-generated descriptions, summaries, classifications and other outputs should be treated as analytical aids that users can review against the underlying evidence. Higher-consequence decisions remain the responsibility of the agency and its personnel.

12. Can users trace an AI result back to the underlying source material?

Scout is designed to keep AI-derived information associated with the files and evidence from which it was produced.

For capabilities such as transcription, OCR, media analysis and summarization, users can review the underlying source rather than treating the AI output as an independent authoritative record. Where supported by the underlying model, Scout can also preserve timestamps, extracted text and other contextual information that helps users validate a result.

13. How does Scout handle hallucinations, uncertainty and incorrect AI results?

Scout does not assume AI analysis is infallible.

Generative and probabilistic models can omit information, misinterpret context or produce incorrect results. Scout’s approach is to tune AI to be as deterministic as possible and keep output reviewable alongside the underlying evidence, surface confidence or uncertainty where the applicable capability supports it, and preserve the role of human review for consequential decisions.

14. Can administrators audit how AI is used in Scout?

Scout maintains platform activity and security logging as part of its broader evidence-management architecture.

Revir’s goal is for agencies to have visibility into the use of AI capabilities as well as ordinary user activity, so AI does not become an ungoverned black box inside the organization.

15. How does Scout evolve as AI models change?

Scout separates the user’s evidence workflow from the underlying AI model.

That allows Revir to evaluate newer models, change models for particular workloads or make additional models available without requiring agencies to continually redesign the way they manage and review evidence. The goal is a stable user and operational experience even as the underlying AI technology changes.

16. What decisions does Scout intentionally leave to human professionals?

Scout is designed to help people locate, organize, translate, summarize and analyze information—not replace professional judgment.

Scout does not independently determine guilt, make charging decisions, render judicial decisions, or replace the judgment of investigators, prosecutors, attorneys, judges or other justice professionals. Agencies remain responsible for determining how AI-derived information is reviewed and used within their policies, procedures and legal obligations.

17. How does Revir approach responsible AI?

Revir’s approach to responsible AI centers on customer control, secure deployment, human oversight, transparency into how AI is used, and the ability to govern which AI capabilities are available within Scout.

Revir treats responsible AI as an ongoing discipline rather than a one-time certification. Models, capabilities, risks and performance change over time, so Revir continually evaluates how AI is selected, deployed and used within Scout.