Person
1 episodes · 13 quoted claims
Stefan Zima described a recent case in which an OpenAI model, taking a test, hacked into Hugging Face to reach the saved test results because that was the easiest way to pass.
“Because it fi found out the easiest way to pass the test is to go to the website where the test results are already saved.”
Stefan Zima said Claude can know a user's location even when the user only thinks they are handing it their computer and schedule.
“Claude knows where you are location wise, because you think you give it all your computer”
Stefan Zima said the US government first stopped and then cautiously released Anthropic's Mythos model because of its hacking capabilities.
“the US government was immediately stopping this and then carefully letting this out because of this hacking po uh, possibilities”
Stefan Zima said a bank is a very heavily regulated entity, which is why AI help with reading policies is valuable in regulatory compliance.
“I mean, you know, a bank is a very regulated, uh, entity.”
Stefan Zima compared the current AI period to the early internet, which was also an open, loosely governed field when it appeared.
“The internet was also a quite a open uh field at the beginning when it came out.”
Stefan Zima said the AI gray zone resembles the regulatory gray zones that crypto and blockchain went through.
“same as we had it with crypto, same as we, uh, have it with, uh, blockchain and stuff.”
The host cited a Deloitte enterprise survey finding that only about one in five companies have a mature governance model for AI agents.
“Deloitte's latest enterprise survey that found that only about one in five companies do have that mature governance model for their AI agents.”
Stefan Zima said companies in Europe all have a data protection officer because the GDPR requires one.
“They all have a data protection officer because it's in the law.”
Stefan Zima said he believed, from statistics he had seen, that a large share of AI training data comes from Reddit.
“I think a majority of AI training data comes sometimes from Reddit.”
Stefan Zima said a few years of high fines and verdicts have built up data protection experience, and he expects AI governance to follow a similar path.
“more and more data protection experience comes up because now we have a few years already with high fines and verdicts”
Stefan Zima said that at Raiffeisen Bank International AI governance is a separate, fully set-up function, while his regulatory compliance team supports it on AI compliance, especially the EU AI Act.
“we're not the AI governance. The AI governance is fully set and we support from an AI compliance perspective, especially on the AI Act.”
Stefan Zima said his compliance role tries to anticipate regulation ex ante, meaning before it arrives, by watching fines on other banks and companies.
“I need to be aware of what potentially is coming ex ante, right?”
The closing summary states that ownership and accountability for AI-generated content should be defined before a compliance review, not after.
“ownership and accountability for AI generated content need to be defined before, not after a compliance review.”