I think the distinction for both topics is important, but we found in the end there is a large overlap when in comes to the controls protecting AI safety and security. We implemented in my company an approach with two teams for AI security and safety. AI security is part of the IT security organization. Both teams need to work together very closely and the resulting integrated AI development framework covers security and safety. I think that would be the recommended approach.
Thanks for your comment, that's really interesting! :) Did your organisation encounter any resistance to this, and how has it been managing cooperation between the two teams?
Actually, we have not encountered any resistance. The safety considerations and approach came out of the team developing the first AI applications and eventually the AI platform. Your medium articles got us started with the structured approach for a AI security. After red team hacked one of the early prototypes there was no discussion anymore. I would say the combination of people really experienced in AI development and thus responsible for the safety plus the mindset of IT security is a perfect fit to set up a AI security and AI safety properly.
AI is a technique. People think of it as life but financially that is off limit. That being said, I don't think that AI is bigger than CS. The way I think about it, is AI is the librarian while computer science is the library.
I think the distinction for both topics is important, but we found in the end there is a large overlap when in comes to the controls protecting AI safety and security. We implemented in my company an approach with two teams for AI security and safety. AI security is part of the IT security organization. Both teams need to work together very closely and the resulting integrated AI development framework covers security and safety. I think that would be the recommended approach.
Thanks for your comment, that's really interesting! :) Did your organisation encounter any resistance to this, and how has it been managing cooperation between the two teams?
Actually, we have not encountered any resistance. The safety considerations and approach came out of the team developing the first AI applications and eventually the AI platform. Your medium articles got us started with the structured approach for a AI security. After red team hacked one of the early prototypes there was no discussion anymore. I would say the combination of people really experienced in AI development and thus responsible for the safety plus the mindset of IT security is a perfect fit to set up a AI security and AI safety properly.
This :)
AI is a technique. People think of it as life but financially that is off limit. That being said, I don't think that AI is bigger than CS. The way I think about it, is AI is the librarian while computer science is the library.