AI will follow a “Concierge” model (graph of AI agents from sophisticated “policy/strategy smart” AIs down to data privacy and comms protocol specialists.
“Data Sharing” – Sharing copies of personal data with 3rd parties who retain that data long term will move to - Person/Individual stored (and/or) controlled “On Demand, purpose specific disclosure (need to know) basis” ideally retained only for an individual interaction/session. Why?
cybersecurity risks/costs (e.g., insurance), “single source of the truth” about the person
Rich personal data built with/by the user for their own control/selective disclosure will displace Surveillance derived (inaccurate, incomplete) data.
AI Agents & Personal Data
Personal data must be shareable (vs. horded by individual AI agents) under user control
Agenda Items and Notes (including all relevant links)
Time
Agenda Item
Lead
Notes
5 min
Start recording
Welcome & antitrust notice
Introduction of new members
Agenda review
Chairs
Antitrust Policy Notice:Attendees are reminded to adhere to the meeting agenda and not participate in activities prohibited under antitrust and competition laws. Only members of ToIP who have signed the necessary agreements are permitted to participate in this activity beyond an observer role.
New Members:
IIW Topics (see above)
Key Discussion Highlights:
Neil Thomson shared insights from the conference, discussing AI architecture in travel applications and data interaction concerns.
Carly Huitema discussed the use of OCA (Object Capabilities for Agents) in government applications, specifically highlighting the use by the BC and Swiss governments. She elaborated on the architecture of OCA and its advantages, such as embedded self-addressing identifiers and feature-based serialization which enhances schema governance.
Issues Raised and Solutions Proposed:
Data Sharing and Privacy Concerns: Neil highlighted concerns related to data sharing and the risks of data breaches, emphasizing the increasing costs of cyber insurance.
Solution: There was a discussion about using federated queries to manage data privacy effectively, ensuring data doesn't leave its original location unless necessary.
Technical Agreements:
Use of JSON Schema for Defining Data Structures: Agreed on the utility of JSON Schema for data structure definition, with Neil and Carly discussing integration methods for current systems.
Handling of General Preferences vs. Specific Context Extensions: There was a consensus on the need to differentiate between general preferences and context-specific data extensions, ensuring tailored data handling according to use cases.
Next Steps:
Both attendees agreed to share and store their presentations on a shared drive for further reference and integration into ongoing projects.
Carly plans to update and elaborate on the use cases and benefits of OCA based on the feedback received at the conference.
5 mins
Review decisions/action items
Planning for next meeting
Chairs
Screenshots/Diagrams (numbered for reference in notes above)