The source is a commentary by Claire Leibowicz and Madhulika Srikumar, published by MIT Technology Review and picked up here via t3n. According to the text, EU regulators and US authorities demand that developers and operators of AI agents monitor them in real time.
What the commentary describes
The authors write that it is currently difficult to determine which AI agents did what, and whether they carried out unauthorized actions. As examples they cite incidents involving agents from OpenAI, Anthropic, Google and Meta. As an analogy, they describe a bank that hires 1,000 new staff without traceable badges and approval chains.
Classification
If you operate AI agents in your company, the demand for real-time monitoring could become relevant to you. Whether and to what extent this applies to your organisation depends on the specific regulation and your field of use; the commentary itself names no binding deadlines or obligations.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
