BARC Data Culture Podcast – „Responsible AI at ZEISS: Compliance Without Slowing Innovation"
The topic hit a nerve, that’s clearly demonstrated by over 18,000 YouTube views in the first three months after publication. That number is not counting the other podcast platforms like Spotify and Apple Podcasts.
I’m talking (or writing) about the episode 191 of BARC's Data Culture Podcast, since renamed the Data & AI Culture Podcast, aired on 27 April 2026: Responsible AI at ZEISS: Compliance Without Slowing Innovation, with host Carsten Bange, CEO of BARC and Florian Biegelmaier, BARC’s data expert.
Our conversation covers what the ZEISS Responsible AI Office actually does inside a global, federated organization. EU AI Act implementation. Risk management. AI security. Why data governance and AI governance need to be thought together. The concrete steps to build policies, training and scalable controls without slowing innovation.
Two lines from the episode carry the argument.
"We're not only asking whether using AI is legally allowed – we're also asking whether it's the right thing to do ethically and socially."
"For us, a data marketplace was like a key turning point because it gives transparency what data assets are already there."
Three takeaways from the conversation
1. AI governance and data governance are kind of the same problem.
An AI system without a data governance foundation is a governance gap that ships. You cannot audit training data provenance if nobody knows which data assets exist, who owns them, or what they were originally cleared for. Data governance is the load-bearing wall a Responsible AI Office builds on top of. Treat them as two projects and you will pay for the same problem twice.
2. Compliance is the floor, not the ceiling.
The EU AI Act tells you what is legally allowed. It does not tell you whether the thing is the right thing to do. A Responsible AI Office that holds both questions in the same review gets to a decision that survives contact with a customer, an employee representative and a regulator. Holding them in separate reviews, or holding only one of them, will look efficient right up until the moment it does not.
3. Transparency of data assets is what unlocks scale.
Governance stops being a bottleneck when the underlying data is discoverable. At ZEISS, a data marketplace shifted the question from "who has permission to ask" to "who has the data, and under which conditions can it be used". Same governance rules. Different friction curve. Scalable controls only work when the thing being controlled is visible in the first place.
Why listen
Most Responsible AI conversations stall at principles. This one moves from principles to operating model: what a global company actually builds when it decides that governance and speed are the same design problem, not opposites. The frame applies to any organization sitting at the same interface. Regulated industries. Federated structures. Multiple product lines. AI already deployed in more places than the central team can name.
If any of these questions are open at your organization, the episode is worth the commute. Is our governance staffed for scale, or staffed for audits. Is our data governance ready for the AI cases we have already promised. Are we treating "legal" and "right" as the same review, or as two meetings that never sit in the same room.
Which review at your organization is currently only asking "is it legal", and what changes if you add "is it the right thing to do" to the same survey you fill or meeting you attend?
Resources
Listen on Spotify, on Apple Podcasts, or watch the episode on YouTube.
Episode landing page with show notes: barc.com – Responsible AI at ZEISS.