/
AI Summary 07 May 2026

AI Summary 07 May 2026

Quick recap

Chris and Noreen discussed the challenges and responsibilities associated with autonomous AI agents, focusing on issues of attribution, responsibility, and the implementation of guardrails to manage AI behavior. They explored methods for detecting AI agents within systems, including monitoring response times, identifying patterns, and using gateways or proxies, while acknowledging the difficulties in tracking external AI products. The conversation also touched on historical experiences with early online systems and the evolution of technology, before concluding with plans to continue the discussion in two weeks.

Next steps

Chris

  • Add more details around how guardrails can work to the relevant document, with the aim of developing solid recommendations.

  • Attempt to drum up more interest in the group/project before the next meeting.

Noreen

  • Add notes to the current document (presumably document 6 or related working document) before the next meeting.

Collaboration

  • Chris and Noreen: Reconvene in two weeks to continue the discussion and review updates.

Summary

AI Agent Monitoring and Control

Chris and Noreen discussed the challenges and responsibilities associated with autonomous AI agents, emphasizing the need for guardrails to control both input and output. They explored methods for detecting AI agents, including monitoring response speed and identifying repeated patterns from specific IDs. The conversation also touched on the potential for a registry system to track AI agents created by employees, though challenges in detection remain.

Technical System Access Challenges Discussion

Chris and Noreen discussed technical challenges related to port numbers and system access, particularly for external products like Claude and Copilot. Chris mentioned that gateways and proxies are being considered to address some of these issues. The conversation then shifted to sharing personal experiences, with Noreen recounting her work at Pricewaterhouse in the 1990s, including her role in a merger project that was completed years after she left.

Industry Group Collaboration Updates

Noreen and Chris discussed their involvement in various industry groups, with Noreen mentioning her work on the Software Safety Specifications panel at Internet Safety Labs and voice interaction projects with W3C and Linux. Noreen shared her experience working with Eve Mailer on a user research project for an ISO membership system test pilot, which ultimately wasn't implemented due to ISO's transition to focus on standard development rather than membership services. The conversation touched on identity-related work, including a project involving Okta system access and identity sharing agreements.

AI Security Discussion Meeting

Chris and Noreen discussed AI security concerns, particularly regarding OpenClaw and other AI agents that can access data on computers. Noreen explained that user research field creates isolated agents that don't leave the computer, and mentioned AI auditing as a way to verify AI behavior. The conversation ended with Chris suggesting that Noreen add notes to their shared document and reconvene in two weeks, with the hope of generating more interest in their project.