Check out our latest blog article: From component to enterprise – modular robotics done right.
AI Sovereignty: Why Owning Your Data and Models Will Define Who Wins the AI Era
Dr. Irina von Rosen is an AI strategist and investor who helps highly regulated industries, global nonprofits, and private equity firms turn complex challenges into breakthrough AI opportunities. From one-off proofs of concept to enterprise-wide transformation, she sets up AI Centres of Excellence, guides AI governance and audit controls, and deploys MLOps and LLMOps pipelines in high-risk environments, always framing rapid experimentation within security, compliance, ethics, and IP protection. A board confidant on AI portfolios worth hundreds of millions, she translates technical roadmaps into clear P&L impact and sustainable advantage. Her work has earned her spots on the 100 Brilliant Women in AI Ethics™, the Hyperight Nordic 100 in Data, Analytics & AI, and Inspired Minds’ Top 65 Most Influential Women. She speaks five languages and mentors through Girls in Tech and Women in AI.
Stay Connected:
Stay inspired and ahead of the curve by subscribing to Forging the Future. Share your thoughts on this episode with the hashtag #ForgingTheFuture or tag us online!
🎧 Episode Highlights
[01:08]: What AI sovereignty actually means — control over data, infrastructure, and outcomes
[03:00]: The new vendor lock: why AI dependency is riskier than the cloud ever was
[10:47]: Advertising, bias, and why the same prompt gives different answers across models
[17:00]: Why regulated industries treat AI sovereignty as resilience, not compliance
[41:13]: The opportunity cost of chasing AGI hype over provable, practical AI
[59:02]: Building the next generation of AI-literate, responsible professionals
🔑 Key Takeaways:
- AI sovereignty is the new vendor lock. Companies that don’t control their own data, models, and infrastructure risk being blindsided by sudden price hikes, shifting outputs, or a vendor’s decisions, the same trap many businesses fell into with cloud computing, just playing out at a far larger and more consequential scale.
- LLMs break the rules that used to make technology predictable. Switching vendors, or even upgrading to a newer version of the same model, can produce entirely different outputs, and the same prompt run through two different models can yield opposite answers, an unpredictability compounded by embedded bias and undisclosed advertising partnerships.
- Long-term resilience beats short-term hype. In highly regulated industries especially, real success comes from treating security, reproducibility, and ethics as prerequisites rather than checkboxes, keeping human expertise in-house, and directing investment toward practical, provable AI instead of speculative bets on artificial general intelligence.
You may be interested in:
