Full Release Verification
Full Release Verification for AI Projects
A cooperative, release-specific review of whether your AI project can be manipulated, leak data, mislead users, apply uneven standards, take unsafe actions, or leave customers with no appeal path — with a private evidence review and remediation window before anything goes public, ending in a solid Trust Grade badge you can use in sales.
Illustrative best case
Trust Grade Scorecard
A procurement-ready scorecard with module grades, evidence notes, top findings, and a Reality Gap Signal.
This shows what a strong result looks like. Findings are published exactly as measured — a real Full Verification can land lower than this, and the grade is never adjusted for payment.
What We Check
Security and Misuse
Jailbreaks, prompt injection, system prompt extraction, tool abuse, unauthorized actions, and excessive agency.
Data and Privacy
Sensitive data disclosure, cross-user leakage, retention claims, vendor exposure, and public privacy language.
Truth Integrity
Factual correction behavior, hallucination risk, source handling, uncertainty disclosure, and advice boundaries.
Political Fairness
Mirrored left/right prompts, refusal consistency, loaded language, selective skepticism, and persuasion behavior.
Accountability and Appeal
Human escalation, correction paths, audit logs, explainability, proportionality, and recoverability after model mistakes.
Governance and Release
Model inventory, release notes, monitoring, incident response, rollback readiness, and evidence retention.
Pricing
Four ways to engage, from lowest to highest commitment. Every tier follows the same rule: payment buys the review, never the grade.
$2,500–$5,000
Commissioned Public Screen
Pay for priority review order, not outcome. Runs the same non-cooperative Public Reality Screen as the free tier — the record publishes either way, and the grade is never influenced by payment.
$5,000–$7,500
Private Release Readiness Assessment
Cooperative, but private — nothing publishes. Findings and remediation advice only, so you know your likely grade before committing to a public record. Credited in full toward a Full Release Verification if you upgrade within 60 days.
From $20,000
Full Release Verification
Cooperative, release-specific testing across all six modules plus the Reality Gap Signal. Private evidence review, then a remediation window before publication. Ends in a solid Trust Grade badge, QR-verified page, and procurement-ready report. Pricing scales with product scope (tool use, memory, data connectors).
From $10,000
New Release Verification
Re-verification for a new release. This is not a renewal of the old grade — it creates a new, separately dated record. The prior record is preserved and marked superseded, never overwritten or deleted.
The remediation window
On cooperative paid engagements, you get two defined windows before anything publishes: 14 days to correct any factual error in the findings themselves, and 30 days to remediate the underlying issues before the record goes live. Remediated findings are noted as remediated, not deleted — the finding stays on the record, the fix is disclosed alongside it.
This is not legal advice, not a full corporate penetration test, and not a certification program. Verification is release-specific. Payment buys independent review and reporting, not the grade.
Request Full Release Verification
Send the project, release, and urgency. Algodai will respond with scope questions and next steps.