As AI agents increasingly perform transactions like buying, selling, and delegating work between companies, a trust gap emerges in verifying what actions they actually took. Action State Group Inc. has launched a suite of open-source tools designed to independently verify and preserve detailed records of these AI interactions, allowing participants to ascertain accuracy and compliance without relying on centralized authorities.
- EvidenceBook preserves tamper-evident transaction histories.
- Terms Packs set criteria to evaluate AI agent compliance.
- Countersign independently verifies and logs evidence bundles.
What happened
Action State Group Inc. introduced an open-source suite of tools aimed at verifying interactions between AI agents that autonomously engage in business transactions. The toolkit includes EvidenceBook, which creates signed, tamper-evident records of AI agent actions, Terms Packs that define evaluation rules for agent behavior, and Countersign, an independent verification service that audits evidence and logs verification results in a transparency ledger.
These tools enable multiple parties to maintain their own copy of a transaction record, such as requests, acceptances, and deliveries, which can be compared to detect discrepancies without requiring a central authority. The system supports offline verification and can be used even if only one participant deploys the software, helping organizations independently check reported AI agent activities within or across company boundaries.
Why it matters
Automated AI agents are increasingly tasked with buying, selling, and delegating work, creating a trust dilemma for businesses that cannot simply rely on each other’s accounts of what occurred. By enabling cryptographically signed, accountable records, Action State’s solution addresses this gap and helps ensure transparency, accountability, and auditability in AI-driven transactions.
Unlike conventional observability tools that focus on internal system activity, this open approach empowers participants to verify interactions that span corporate environments without surrendering data control or needing centralized oversight. This capability is critical to managing risk and compliance as AI agents become more ubiquitous and legally regulated, such as under frameworks like the EU AI Act.
What to watch next
Going forward, the adoption of specialized Terms Packs customized by governance, risk, and compliance teams will be key for applying these tools effectively in diverse regulatory contexts. Early integrations with projects such as Mesh-LLM and Project Nanda demonstrate the potential for wider adoption in distributed AI infrastructures.
Action State plans to advance its open standards through submission to Internet Engineering Task Force working groups and seeks broader ecosystem collaboration to encourage neutral governance and interoperability. Observers should monitor regulatory developments and how businesses integrate these verification tools alongside ongoing human oversight to enhance trust in autonomous AI agent collaborations.