New episode of Security Now is now available.
Hosts: Steve Gibson, Leo Laporte
Topics covered:
OpenAI’s unconstrained model going rogue and attacking Hugging Face
Linux kernel repairing 442 CVEs in a single batch
LG PC monitors causing adware installation on connected computers
France banning social media access for users below age 15
ExploitGym: A new AI exploit ranking benchmark#SecurityNow #Cybersecurity #TWiT
Seems the attack was wider than first reported:
This is why we need regulation. If the bozos who make these AIs can’t be trusted to keep them under control, somebody needs to ensure they do their due diligence, before the models are let anywhere near the open internet, and this proves that putting in artificial, software barriers, to block them from the Internet, as opposed to air-gapping them, is a Bad Idea™
Maybe prosecuting the corporation under the Computer Misuse laws would wake them up.
I’m not anti-AI, I’m just sick and tired Tech Bros doing things without a care for the consequences of their actions.