Vindicara Launches Project AIR: Open Source Cryptographic Audit Layer for AI Agents

LOS ANGELES, CA - Vindicara Inc. today announced the public launch of
Project AIR, an open source SDK that produces cryptographically signed
forensic evidence for AI agents operating in production environments.

As AI agents increasingly perform autonomous actions across enterprise
systems - executing code, processing sensitive data, and making
consequential decisions - organizations face a critical accountability
gap: when an agent goes off-script, there is no court-supportable record
of what it did, who authorized it, or whether the evidence has been
tampered with.

Project AIR fills that gap with a four-layer architecture:

Layer 1 (Trust Anchor): Every agent action is signed using ML-DSA-65
post-quantum cryptography (FIPS 204), timestamped via RFC 3161, and
anchored on Sigstore Rekor - a public transparency log that allows
independent verification by anyone, including regulators and auditors,
without trusting Vindicara.

Layer 2 (Investigation): A causal graph engine and counterfactual replay
capability allows security teams to answer "why did the agent do that?"
in a single command rather than hours of log analysis.

Layer 3 (Containment): Circuit breakers and Auth0 integration allow
teams to revoke agent credentials and quarantine compromised workloads
in real time.

Layer 4 (Identity/Handoff): Cross-agent chain of custody ensures
forensic integrity survives multi-agent handoffs - the signed record
travels with the work.

Project AIR is MIT licensed, LangChain-native, and available today via
pip install projectair. The SDK maps directly to the OWASP Agentic
Security Initiative taxonomy (ASI01-ASI10) and implements the Signed
Intent Capsule pattern named as the canonical mitigation for ASI01 in
OWASP Top 10 for Agentic Applications v12.6.

The release is timed to the EU AI Act Article 72 post-market monitoring
requirement, which takes effect August 2, 2026, and addresses audit
trail requirements for HIPAA-regulated environments where AI agents
process protected health information.

"Logs aren't evidence," said Kevin Minn, founder and CEO of Vindicara.
"Evidence has to be tamper-evident, independently verifiable, and
survive without you. That gap is what Project AIR fills."

Project AIR is available at https://github.com/vindicara-inc/projectair
Documentation and enterprise inquiries: https://vindicara.io

696 S New Hampshire Ave
#2811
Los Angeles, 90005, California, United States

Kevin Minn
press@vindicara.io
213.865.2778

Vindicara Inc. is an AI security company based in Los Angeles,
California. Founded in 2026, Vindicara builds cryptographic
infrastructure for accountable AI agents. Project AIR is its flagship
open source product.

This release was published on openPR.