Urgent.News

What's breaking now, across thousands of outlets.

AI

Startup Action State launches tools for verifying interactions between AI agents

Action State Group Inc. is launching a set of open-source tools designed to let businesses independently check records of what artificial intelligence agents did, addressing a trust problem as automated systems begin buying, selling and delegating work across corporate boundaries. The offering combines EvidenceBook, which preserves tamper-evident records, Terms Packs for evaluating agent behavior…

Startup Action State launches tools for verifying interactions between AI agents

Startup Action State Group Inc. has introduced a suite of open-source tools called EvidenceBook, Terms Packs, and Countersign to help businesses independently verify the actions of artificial intelligence agents. This initiative aims to address the trust issues arising as AI agents increasingly engage in corporate transactions like buying, selling, and delegating work.

The core of the offering is EvidenceBook, which creates tamper-evident records of AI agent activities. Buyers can log their requests and payments, while providers document what they delivered. Comparing these records can reveal discrepancies without needing a central authority. Both parties retain control over their information and can share additional details during outcome evaluations. The verification process works offline and is free of charge.

Another key tool, Terms Packs, allows parties to define evaluation criteria and methods based on regulatory or business rules. These packs outline which requirements were met, missed, or could not be evaluated, but they don't certify legal compliance. They require human review for implementation.

Countersign is a verification service that checks specified properties of an evidence bundle, signs its findings, and records them in a transparency log. However, it doesn't confirm if the claimed outcomes actually happened or if the evidence is complete.

Action State emphasizes that these tools verify events occurred as reported, but they don't guarantee the accuracy of the content or the completion of captured actions. The company also mentions that human review plays a crucial role in the process, suggesting daily assessments by AI judges followed by periodic human reviews of samples.

Initial integrations include Mesh-LLM, an open-source distributed inference network where nodes maintain signed records and can compare exchanges with peers. The system also preserves evaluation results, enabling auditors to examine the underlying evidence. Action State plans to seek neutral governance for the specifications and implementations under the Apache 2.0 license. They expect to generate revenue from independent verification services once there are relying parties that accept the verification results.

Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at siliconangle.com →

More in AI

More from Thursday 22 October →