Private evaluation
Evaluate SentraMind inside explicit boundaries
This preparation guide explains who an evaluation is for, how it would run, what each party would provide, and what evidence to expect. Private-evaluation intake is not yet enabled. This page collects and transmits no information.
Intake is not yet enabled
We are finalising accountable ownership, privacy notice, retention, hosting and security approval, and a monitored first-party contact channel. Until those controls are approved, we will not publish a form, contact address, or response-time promise.
Who an evaluation is for
- Organisations validating AI-use controls inside a local boundary
- Teams able to nominate security, platform, privacy or compliance owners
- Design partners prepared to begin with synthetic or internally approved material
The current evaluation boundary
Core and Sentry are available for POC. Cartographer is coming next. The browser extension, endpoint agent, Active Directory integration, classifiers and remaining crew are design-partner roadmap items. Browser demonstrations explain behaviour; they are not proof of production readiness.
Proposed evaluation stages
- 1
Qualification and readiness
Confirm objectives, decision owners, environment constraints and acceptance criteria.
- 2
Architecture and data boundary
Record deployment location, network boundaries, log handling and approved data classes.
- 3
Corpus preparation
Prepare synthetic or internally approved examples; upload nothing to this website.
- 4
Local installation planning
After readiness and security review, agree an isolated environment and installation procedure.
- 5
Baseline and tuning
Begin in observe mode; record rules, exceptions and known blind spots.
- 6
Observe and warn
Compare detections and false positives before changing user experience.
- 7
Enforcement decision
Discuss enforce mode only with explicit authority and documented exit conditions.
- 8
Report and next step
Summarise evidence, limitations, residual risks and the decision to continue or stop.
Responsibilities
Evaluating organisation
- Nominate business, security, privacy and technical owners
- Provide synthetic or internally approved material and environment facts
- Approve access scope, change windows and exit conditions
- Decide whether results meet internal risk and compliance requirements
SentraMind evaluation team (once intake opens)
- Record scope, assumptions and product maturity
- Help define installation, configuration and test plans
- Explain detections, limitations and reproducible steps
- Prepare a report without replacing the customer’s risk decision
Measures and deliverables
The plan should define, by detector class, the sample set, expected and observed results, false positives, false negatives and untested scope. Proposed deliverables are a scope and architecture record, configuration and decision log, detection summary, limitations register and next-step recommendation. Throughput, latency and model-quality numbers must come from the actual evaluation; this page promises none.
Limitations that remain in force
- An evaluation is not certification, legal advice or production approval.
- Some interfaces and demonstrations are behavioural stand-ins, not a connected full product.
- No non-public material should be exchanged until handling, retention, access and deletion rules are documented.
- Any NDA and data-handling terms must be separately agreed before non-public material is exchanged.
What you can do now
While intake is closed, run the registration-free browser demonstrations and review the current product boundary.