Submit topics for agenda - Aug 27

[jpanat]

Agenda

1. 2 new blogs from Jim and Jean
   1/ https://blogs.agntcy.org/technical/identity/2026/08/25/agent-identity-tbac-a2a-task-authorization.html
   2/ https://outshift.cisco.com/blog/ai-ml/ai-security-for-autonomous-agents

2)A2A <-> AGNTCY gap
3) ServiceNow Use case
4) D&B VC prop: Raj

[jpanat]

Meeting Recording:https://zoom.us/rec/play/1fx2jgQZW2yIT1cGXmllZA5V9t6h8SZ5yZDE35O_fTIpKOzTIJthFZIwEvXUMD8nbPIC01fwY2PvU87g.jKVZB0ZN9Ebs2Q3L?accessLevel=meeting&canPlayFromShare=true&from=share_recording_detail&continueMode=true&oldStyle=true&componentName=rec-play&originRequestUrl=https%3A%2F%2Fzoom.us%2Frec%2Fshare%2FgHbCYIZO9TSKbNcRGRklAj0BHaGL8Ku6Rr93bx0BNXp4TX_lbyBY2f57lvNZTckR.nWFOdxastz-gRxnp

[jpanat]

Quick Recap
The meeting focused on updates and discussions around agent identity, security, and integration within the Agency and broader Linux Foundation ecosystem. JULU highlighted recent blogs addressing OpenAI and Hugging Face issues, and encouraged group members to amplify these messages. Jean presented an analysis on aligning Agency Identity Service with A2A, discussing gaps and potential integrations. Sarah suggested leveraging existing architecture patterns and use case groups to address security gaps, rather than forming new working groups. Rajagopal from DNB introduced a proposal for verifiable credentials for agents, sharing details on their POC and schema, and discussed potential collaboration with the group. Tomer announced the approval of an OWASP workstream for an Agent Identity Maturity Model, aiming to guide organizations on best practices. The group also discussed cross-group collaboration, demos, and the role of policy enforcement in agent interactions. Various participants shared updates on ongoing projects, demos, and potential areas for further exploration in identity and security for AI agents.

Next Steps
All attendees: Promote and share the two published blogs (Jim's and John's) on LinkedIn, Twitter/X, etc.
1/ https://blogs.agntcy.org/technical/identity/2026/08/25/agent-identity-tbac-a2a-task-authorization.html
2/ https://outshift.cisco.com/blog/ai-ml/ai-security-for-autonomous-agents

Manish: Share the draft of the Gateway blog with the group for review and publication support.
JULU: Post the links to the two blogs in the meeting chat for access.
JULU: Discuss with John (Parello) about aligning the A2A gap-closing effort with the use case group and AAIF.
JULU: Set up a meeting with the larger ServiceNow leadership group, then initiate a smaller ServiceNow Identity work stream including Sarah, Amrita, and John.
Rajagopal: Share the sample W3C VC schema with the group, preferably in the Slack channel.
Jean, Sri, Rajagopal: Collaborate to work on the schema integration between D&B's verifiable credentials and the Agency Identity Service.
Tomer: Announce the OWASP Agent Identity Maturity Model work stream, create a Slack group, and share the draft document for collaboration.
All interested parties: Provide feedback on the brokered IDP implementation proposal (GitHub discussion 18) within a week.
agntcy/agent-identity-demos#18

Summary
AI Platform Security Working Group:
Julu led a meeting of the working group, discussing recent blog posts from Jim and John addressing challenges with OpenAI and Hugging Face, and opportunities for identity and access control in AI platforms. He encouraged participants to promote these blogs on their social media channels and informed the group that support is available for writing and publishing thought leadership content. The conversation ended with John preparing to present an analysis of gaps within the agency, suggesting potential opportunities for a working group to address these issues.

TBAC-AIS Integration Planning Discussion:
Jean presented work on integrating TBAC and Agent Identity Service with A2A version 1.0.1, highlighting similarities in user approval processes and task structures. The team identified potential implementation gaps and discussed opportunities to incorporate these ideas into current POCs. JULU suggested creating a working group to address the integration challenges, and Sarah mentioned considering how composable agency pieces fit into broader security architecture patterns.

Architecture Workflow Collaboration Updates:
Sarah and JULU discussed the evolution of architecture workflows and use case patterns, including the transition of the Andor group to the Agentic AI Foundation (AAIF). They explored the potential for aligning these patterns with the OpenAI Hugging Face workflow and identified the need for cross-foundation collaboration. JULU provided an update on the ServiceNow project, noting progress on a proof of concept involving agent identity and interoperability between IDP providers. The discussion concluded with plans to involve Raj from DNB to present a proposal on verifiable credentials.

Agent Verification System Development Proposal:
Rajagopal discussed a feature request submitted to the Identity GitHub repo regarding the need for verifiers for autonomous agents, particularly in the context of business certifications. He explained their proposal to issue W3C VC credentials containing DUNS identifiers and basic demographic information, with plans to collaborate with agencies like DMV to create a more workable ecosystem. The team has developed APIs and websites for publishers to authenticate and claim ownership of agents, with D&B hosting revocation lists and credentials with expiry dates. Rajagopal proposed creating a working group to further develop the schema and determine how to integrate with the OASF standard, allowing verifiers to filter agents based on both skills and certification status.

Agent Identity Verification Processes:
The team discussed agent identity verification processes and integration approaches. Rajagopal explained the current DNB onboarding portal system for verifying agent identities, including domain ownership validation through DNS records or well-known endpoints. The group agreed that Jean, John, and Sri would collaborate with Rajagopal on developing a schema for the verification system. Tomer announced the creation of an OWASP work stream focused on developing an Agent Identity Maturity Model to help organizations assess their agent identity tools and configurations. The team also discussed ongoing work on a white paper documenting modular identity service implementations across CNCF, Agent, and AAIF stacks, with Manish seeking feedback on a proposed brokered IDP implementation approach.