Scope. Safety. Evidence. Accountability.

How Balhence handles testing and client data

This page states the operating standards behind a Balhence engagement. The signed scope and agreement remain controlling for a specific project.

Before access

Written authorization defines the boundary

No testing begins from a form submission or an informal request. The parties first agree the systems, environments, identities, methods, schedule, evidence handling, escalation contacts and stop conditions in writing.

Third-party assets require permission from their owner. A change in scope, environment or safety conditions pauses the affected work until the boundary is confirmed again.

Minimum safe proof

Testing is controlled, observable and proportionate

Evidence should prove the issue without creating unnecessary impact.

01 / IDENTITIES

Controlled accounts

Use dedicated test identities, synthetic records and agreed tenant boundaries where practical.

02 / LIMITS

Rate and action controls

Set testing windows, request limits, prohibited actions and an escalation route before execution.

03 / IMPACT

Smallest useful demonstration

Stop when the agreed evidence establishes realistic impact; do not collect unrelated data.

04 / RESPONSE

Urgent notification

Use the agreed contact path when a confirmed issue needs attention before the report is complete.

Client material

Data handling is agreed for the engagement

The project agreement identifies what material is needed, how it can be transferred, who can access it, where it may be processed and when it should be deleted or returned.

Handling principles

  • Collect only the material needed for the agreed work
  • Keep credentials and tokens out of ordinary findings and email
  • Redact secrets, unrelated records and personal data from evidence
  • Share sensitive material through an agreed channel
  • Limit access to people involved in delivery

Project-specific decisions

  • Approved storage and processing locations
  • Source-code or repository access
  • Evidence retention and deletion date
  • Subprocessors and approved AI services
  • Customer, regulator or auditor handling requirements
Approved assistance only

AI supports the work; people remain responsible

AI may support approved research, analysis, code, tooling and documentation. Its use does not expand authorization, remove a safety gate or turn an observation into a delivered finding.

No public model receives client credentials, tokens, customer records, source code or unredacted vulnerability evidence unless the client explicitly approves the provider, purpose and processing terms. No model autonomously tests production.

Read the detailed human-validation approach.

Report quality

A delivered finding has a human review trail

A candidate observation is checked before it asks your team to act.

Reproduce

Confirm the behavior in the agreed environment with clean controls where practical.

Bound impact

Explain realistic impact, affected roles or tenants, preconditions and known limits.

Prepare evidence

Provide useful reproduction steps and redact secrets or unrelated data.

Re-test clearly

Record the tested version and whether the issue is fixed, partial, open or not re-tested.

See the format in the fictional sample report.

What this page does not claim

Trust comes from precise statements

Balhence does not describe itself as certified, 24/7 monitored, vulnerability-free or guaranteed to find every issue. An assessment reflects the agreed scope, access, environment and time available.

Framework references explain methodology; they are not certifications or endorsements. A report can support an audit or customer review, but the relevant reviewer decides whether it meets their requirement.

Responsible disclosure

Found a security issue in Balhence?

Use the published security.txt contact and guidance. Do not include passwords, private keys, session tokens or unrelated personal data in the first message.

For questions about a proposed engagement, contact contact@balhence.com or send a scoped enquiry.