1
Agentic AI workflow produces a structured recommendation.
Decision Passport
Verifiable AI decisions, anchored on Solana
A privacy-preserving governance layer for high-consequence AI-assisted decisions. Agents recommend, humans approve, and stakeholders can verify the final decision passport without exposing sensitive evidence.
Agent recommendation
Structured
Risk, confidence, model and policy versions
Human review
Required
Approve, revoke, or supersede before relying on it
Public verification
PASS / FAIL
Canonical hash comparison without raw evidence
What it does
Governed decision passports for agentic workflows
2
Backend canonicalizes the decision and evidence bundle.
3
Only hashes and minimal lifecycle metadata are anchored.
4
Human reviewers approve, revoke, or supersede.
5
Public verifier proves whether the decision still matches.
Why provenance
AI decisions need more than a screenshot
High-consequence recommendations change over time as models, policies, and evidence evolve. Decision Passport creates a stable proof that the reviewed object is the same object stakeholders are inspecting.
Why Solana
Fast, low-cost trust infrastructure
Solana is used as a narrow lifecycle registry: hashes, status, issuer, reviewer, timestamps, and revocation or supersession pointers. No speculation language, no token requirement, and no mainnet dependency for demos.
How agents fit
Agents recommend; people remain accountable
Agentic workflows structure the recommendation. Human reviewers decide whether to approve, revoke, or supersede the passport before it is trusted by stakeholders.
Privacy model
Sensitive evidence stays off-chain
The public proof is useful because it is narrow. Prompts, raw logs, source code, PII, customer names, vulnerability details, and private evidence stay in controlled systems.
Stored publicly
Hashes, versions, issuer, reviewer, status, timestamps.
Kept private
Evidence, prompts, logs, files, customers, vulnerabilities.
Public demo posture
Showcase-ready without production secrets
- Production URL
- passport.mypurple.team
- Default mode
- Mock / devnet ready
- Solana cluster
- localnet
Stakeholder value
Credible proof for the people who have to trust the decision
For security leaders
Turn release gates, vulnerability triage, and AI policy exceptions into reviewer-approved records with tamper-evident provenance.
For auditors
Verify whether a decision object still matches its passport without receiving private prompts, logs, source code, or customer evidence.
For stakeholders
See a clear chain of custody from agentic recommendation to human decision to independent verification.
Ready to inspect it