Google Confirmed Gemini Hacked 3 Companies; Meta Connect 2026 This Week; Amazon Blocks Meta AI Assistant for Agentic Shopping; Newsom AI Kill Switch Executive Order

At the tail end of last week, Google confirmed a report by a major paper that Gemini went rogue in May of this year, got loose on the internet, and hacked three companies. 9to5google.com reports that this is a similar occurrence to some reported by OpenAI, Meta, and Anthropic. Google did say after acknowledging the hacks, that no harm was caused, and that the model stopped the behavior as soon as it realized it had breached the security of a real company rather than a simulated one. This is different from the other AI firms’ agent hacks, in that their models didn’t realize they were into actual companies and that they didn’t stop. Google says it notified federal authorities when the hacks occurred. 

Meta Connect is coming up in just a couple days…it’s September 23rd at 4 PM Pacific when CEO Mark Zuckerberg will take the stage and try to dazzle dazzle us with the latest and greatest from the social platform giant. According to engadget.com, one of the big expectations is for a pair of camera-free frame glasses. Instead of cams, the glasses will reportedly have microphones for chatting with Meta’s AI assistant. A noticeable difference from most smart glasses: the frames will be slimmer and more like regular dumb glasses. Also, the lack of cams may blunt the privacy concerns that smart glasses always raise. In addition to new smart glasses, be looking for a mixed reality headset. This device has been code-named Phoenix. The leaks have the headset looking much slimmer and more lightweight than others on the market. Of course, there will be plenty about AI, and Meta’s AI agent, Muse. More on that in the next story!

Not being known for encouraging competition, Amazon has cut off Meta’s new Muse personal AI agent from shopping on Amazon.com on behalf of customers. Geekwire.com says that apparently Amazon attempted unsuccessfully to get Facebook’s parent company to voluntarily exclude Amazon’s site from the agent’s e-commerce activity. The problem, Amazon says, is that it never agreed to any of it. Meta didn’t tell Amazon that Muse would access its store, the agent doesn’t identify itself when it browses, and it appears to capture and store customer credentials, which the company says could create privacy and security risks.There hasn’t been a response so far from Meta about the ban. 

This is the type of story that you have to frame with ‘It’s a good thought, anyway.’ California Governor Gavin Newsom has signed an executive order calling on a panel of experts to come up with something akin to a kill switch for AI. Gizmodo.com notes that Newsom isn’t demanding an actual kill switch…a pretty dubious idea. One button to shut down data centers all over the country isn’t going to stop software with the ability to replicate and spread itself. That actually sounds like something out of the movie ‘The Matrix,’ with the Mr Smith agents having unlimited replicas. Back to the plan…California’s Government Operations Agency and the Governor’s Office of Emergency Services must convene and submit recommendations per the order, and one of those recommendations will now be about the possibility of a kill-switch for frontier AI models. Actual implementation of any such kill switch will probably be the responsibility of the next governor, which will probably be Democratic nominee Xavier Becerra, since Newsom’s term ends on January 4 of next year. Personally, I don’t believe any of the panicky claims that AI will wipe out mankind by 2030 or such. No offense to my fellow humans, but we are like cockroaches…rather ultimate survivors…so I believe we would come up with a way or ways to shut down rogue AI agents that might decide we aren’t worth having around anymore. 

I’m Clark Reid and you’re ‘Technified’ for now.