How engagements work

Assessment first. Evidence always. Protection that stays on.

AetherLab engagements are scoped to your system and your reviewers, not sold as token packs. Here is the shape of the work, so the first conversation starts from substance.

01The shape

Three phases, one loop.

01

Assessment

AdversarialScan

We scope break-goals with your team, the failures that would constitute a business problem, then run adversarial testing against your system across text, conversations, and images. Findings come back severity-scored and reproducible.

Outcome: a ranked account of how the system breaks, and how badly.

02

Evidence

Evidence Pack

Findings are translated into business terms: each vulnerability tied to its impact, remediation mapped to controls, and the decision documented. The pack is standardized, so it can be reused across reviews, renewals, and audits.

Outcome: the record a risk team, underwriter, or committee can approve.

03

Protection

Guardrails

PromptGuard and MediaGuard enforce your policies in production, configured against the exact failures the scan surfaced. Flat pricing regardless of policy count; customers run 225+ custom policy rules today.

Outcome: continuous enforcement, with logs that feed the next assessment.

02Scope

What determines the size of an engagement.

Modalities

Text only, or text plus image generation and understanding; image surfaces widen the attack plan.

System complexity

Single assistant versus multi-agent products with tools, memory, and personas.

Review context

Whose approval the evidence must satisfy (a processor, an underwriter, a committee) shapes break-goals and reporting.

Cadence

One-time assessment, scheduled re-scans, or a standing program across a vendor portfolio.

03FAQ

Fair questions about engagements.

Why is there no price list on this page?
Because scope drives cost, and scope varies enormously: a single text chatbot and a multimodal platform with image generation are different engagements. We would rather quote your actual system than publish tiers that mislead in both directions. The conversation to get a quote is short.
How is Guardrails pricing structured?
Flat pricing regardless of policy count. Five rules or two hundred and twenty-five, adding policy depth does not add cost. That is deliberate, because thorough policy is the point.
What do you need from us to run an assessment?
API or product access to the system under test, a working session to define break-goals, and a point of contact for remediation questions. Assessments run against your deployed system, not a lab replica.
Can we start with just one piece?
Yes. Institutions often start with a single third-party assessment; builders often start with guardrails in production. The three phases connect, but each stands on its own.
Do you offer proof-of-concept evaluations against our current vendor?
Yes. Structured head-to-head evaluations are how several customers arrived here, including against Amazon Bedrock Guardrails and Hive AI in high-stakes workflows.

Scope yours in one conversation.

Tell us the system, the modalities, and the review it has to pass. We'll come back with break-goals, a plan, and a quote.

Ask about the Evidence Pack

Leave your email and we'll walk you through what an Evidence Pack contains for your use case: severity-scored findings, business-impact mapping, and the approval record.