Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[ai-interest-group] SAIF Meeting notes - Oct 7th 2026
Dear all,

As per Sharon's message from earlier today, we are pleased to celebrate our 18th member, Antonio Jara, joining us for Libelium! It was great talking to you earlier today, Antonio!

Meeting overview

The Sovereign AI Foundation (SAIF) interest group met for a panel moderated by Michael Berns about Technology & trend tracking North American edition, followed by working breakouts.

Panelists were David Gilmore, CEO and Co-founder of Rayline.ai and Workshop.ai, with a previous nine-figure Silicon Valley exit;

Brandon Carl, AI and Product Leader who built AI that processes tens of billions of communications; and our very own community member

Matthew Khouzam, Principal Researcher at Ericsson.

Panel discussion

We heard about the latest trends in terms of TypeSafe / Jev, expanding the tool chain from the LLM sledge hammer as Brandon Carl called it, and had David Gilmore explain the perspective of the startup trying to survive in the market dynamics as well explain some of the tough decisions when designing AI products. Matthew shared great insights from the corporate & compliance perspective in terms of model selection & developer tooling.

The panel covered three themes at a high level:

  • Hyperscaler dependency vs. the cost of lock-in: whether building on closed hyperscaler APIs is a sustainable long-term strategy, the hidden architectural cost and technical debt for enterprises, and the operational challenges of moving AI prototypes into full-scale production.
  • Agentic AI in mission-critical systems: how to bring autonomous agents to risk-averse enterprises, how to balance human-in-the-loop control with performance, and whether latency requirements push agents toward the edge.
  • Sovereign AI and open source: The panel discussed open source as a path away from centralized dependency, whether interest in AI sovereignty extends beyond Europe to North America, and the enterprise readiness of open source AI stacks for core operations.

Audience Q&A and outlook

The panel took closed with a lightning round biggest surprise they expect in enterprise AI architecture over the next 12 months and took questions from the audience.

Attendance and work breakouts

We had 31 participants. Uday Kamath, Chief Analytics Officer at SMARSH and author of eight standard works on AI, was sadly absent, we wish him a speedy recovery and hope to have him join us on a future occasion. At the top of the hour, we moved into the breakout rooms to collaborate on four topics:

  • Security & Safety: user behavior, layered security, trade-offs between security and convenience, and sandboxing.
  • Ethical AI: gaining traction for responsible use, the increasing use of AI in the workplace, keeping or breaking the loop, and the work still being done by people. We discussed the shift toward approval-based workflows and how much to keep the human in the loop.
  • Developer tooling: Edge AI how to turn the discussion into concrete action, starting with an initial first step.

Looking forward to speaking to you again on the 28th of October, when we will start at 430 pm CET. 



Michael Berns


Michael Berns

Head of Artificial Intelligence

Eclipse Foundation


Let's collaborate & talk AI - pick a slot


Mobile   :+49 151 15252798

LinkedIn: aithoughtleader


Want to shape the open-source AI landscape? 

🚀 Collaborate with us at the newly launched Sovereign AI Foundation

🤖  AI Startup Mentor & Board Advisor since 2013

💡 Insights: Explore some of my 100+ AI publications 

Back to the top