5 Comments
User's avatar
Bianca Schulz's avatar

A friend of mine is very deep into this topic from the engineering side and he teaches me his framework. I think it's brilliant and perhaps one of very few ways how to really have AI agents under control. What do you think about this one from a lawyer perspective?

Here is the link: https://biancajschulz.substack.com/p/what-the-hell-is-ontology

A summary can be found in one of my latest notes.

Silvia Stepitova's avatar

Thanks for sharing this, Bianca. The three-way split is what makes this work from a legal perspective. Vocabulary, permissibility, and calculation each auditable independently. That's what the EU AI Act's transparency and logging requirements are reaching for but rarely get in practice. And the 'not less intelligence, but a boundary drawn in advance' philosophy is very close to what meaningful human oversight should look like.

Bianca Schulz's avatar

Thanks for your response! Great to hear that a lawyer likes it too

Mike Schlottman's avatar

Nice call out on the lag between the EU AI Act and the technological reality. Agentic AI has become the One Ring: coveted by leaders chasing power and profit, corrupting them in the wearing. Article 25 is the chapter nobody wants to read, because now the ring is stuck to you.

Silvia Stepitova's avatar

Thank you, Mike. The One Ring analogy keeps getting better the more I think about it. Article 25 really is the chapter people don't want to open.