️ with Sebnem Erener, AI Governance & Policy Expert
Opening
Europe is investing billions in open source AI as the trustworthy
path, and this episode asks a simple question about that plan:
does open actually mean trustworthy? Sebnem Erener works on AI
governance, and her answer is no, or at least not on its own.
Open source spreads access to a model. Whether you can trust it,
and who answers when it goes wrong, are different questions with
different answers.
Episode overview
An open model passes through a lot of hands before it reaches
you. Sebnem counts five or six: uploaded to a platform, hosted
somewhere else, routed, fine-tuned, served. The law usually wants
one company to hold responsible, and that's not how the open
ecosystem works. She spent years at Klarna as its Head of AI
Legal, Regulatory and Ethics, and she now studies where this gap
sits. We talk about why she cares more about trustworthiness than
trust, what open source can and can't fix on its own, and why she
thinks a person can be held liable but a model never can, no
matter how independently it acts. It's a conversation for anyone
building on an open model, or writing the rules for one, who
wants to know who's actually accountable.
What you'll hear
Why an open model can leave you less sure who's liable than a
closed one
The line between real openness and what Sebnem calls "open
washing"
Where Europe should be patient, where it should be bold, and
why one timeline for all of it is a mistake
Why calling AI "agentic" doesn't move the responsibility onto
the machine
The thing Europe is short on, which isn't innovation
About Sebnem Erener
Sebnem Erener works on frontier AI governance and policy across
EU, UK, and US frameworks. Before going independent she was Head
of AI Legal, Regulatory and Ethics at Klarna, so she has handled
these questions inside a company and not only in theory. She's an
advocate for open source AI, and she's precise about what it does
and doesn't solve. Her forthcoming paper on the governance of
open source AI intermediaries runs underneath a lot of this
conversation.
️ Chapters
[00:00] What we mean when we say "trust"
[05:00] The five or six hands on every open model
[09:00] When open makes accountability harder
[12:00] One package, four timelines: patience and boldness
for Europe
[14:30] Why "agentic" lets no one off the hook
Links
Sebnem Erener on LinkedIn -
https://www.linkedin.com/in/sebnemerener/
Sebnem Erener website - https://www.sebnemerener.com
Eva Simone Lihotzky on LinkedIn -
https://www.linkedin.com/in/evalihotzky/
European Commission tech sovereignty package -
https://digital-strategy.ec.europa.eu/en/policies/eu-tech-sovereignty
OpenEuroLLM - https://openeurollm.eu
Related episode - beyond the black box with Prof Dr Tina
Weisser:
https://open.spotify.com/episode/6yKMLQNnDoa6NqqClJrryN?si=3AWXxowURVujJL5m3zD5QQ
Kommentare (0)
Melde dich an, um einen Kommentar zu schreiben.