Who carries the responsibility?
Regulators in the UK and Australia want proof that AI is safe, and CIOs expect to answer for AI they never chose. This week's common thread: you cannot outsource ownership of AI.
Four stories, one message. Accountability for AI does not move to the technology, a vendor or another department. If you deploy AI, you need to be able to show who decides and how it is governed.
-
Information Commissioner's Office (ICO) ·
ICO secures changes from leading AI developers as scrutiny extends to AI agents
The ICO is blunt: an agent acting autonomously is no excuse for poor compliance. If agents touch personal data, the organisation that deploys them stays accountable, so this belongs in the privacy assessment up front. The call for evidence is open until 20 November.
-
ABC News ·
Australia outlines plans to make AI companies prove their safety systems work
Australia wants AI developers to prove their safety systems work, borrowing from banking and aviation oversight, because voluntary codes are not enough. I expect the same burden of proof to reach European boardrooms: it will not be enough to say you control AI, you will have to show it.
-
CIO Dive ·
Who actually controls enterprise AI?
Thoughtworks asked 3,200 CIOs who owns and funds AI and found no dominant model. Yet 94% of US CIOs expect to be held responsible when AI deployed by other business units fails. Put decision rights and accountability with the same person, or the first incident becomes a search for someone to blame.
-
ITPro ·
ISACA CEO says AI compliance will be like 'SOX on steroids'
ISACA's CEO Erik Prusch expects AI compliance to be heavier than SOX, with board members named in lawsuits when AI goes wrong. In Europe the AI Act lands on top of NIS2, DORA and the Cyber Resilience Act. One coherent governance framework beats four separate compliance programmes.