US House of Reps tells staff: No Microsoft Copilot for you! Staff working at the US House Of Representatives have been barred from using Microsoft's Copilot chatbot and AI productivity tools, pending the launch of a version tailored to the needs of government users. According to documents obtained by Axios, the chief administrative officer (CAO) for the House, Catherine Szpindor, …
Seems sensible enough. Not just the risk of potentially leaking classified docs, but also inserting random nonsense into various other missives that aren't checked carefully before being fired off.
There are some "AI" things that are useful, like email spam filters that "learn" and "adapt" to identify likely spam even if it's something it's never encountered before. However, that's about the full extent of what I can think of for useful applications for AI for the average person. We still likely have a good decade plus before we have "AI" that actually does something useful. Something akin to the Star Trek computer, where you can have it process natural language requests that don't require very specific syntax and deal with ambiguity without making up a load of bollocks. Maybe Dell and the few other remaining PC OEMs out there should consider diversifying their businesses more. I know Dell owns EMC, so that's a good start. HP seems to be moving into Printer-as-a-Service and Ink-as-a-Service scams, so that's... something. The days of mass refreshes are over, and the smart CEO would have long ago planned accordingly, which I guess rules out pretty much every CEO at the major OEMs left standing. I suppose in some respects, the model used by Acer and Asus, of buying only enough materials to build X number of a specific model turns out to be the way to do it, even if it means your warranties aren't worth the paper they're printed on. How many people are going to go to the hassle of filing a small claims complaint?
I work in state gov. The highest levels (Directors etc) are now spamming everyone with book-length AI written emails. Literally makes everyone cringe when they come in.
This is basically the use-case for most of this stuff for idiot normies, to write terrible and cringe inducing emails. (the last one was a long list of all the employees who had died this year-"Jim passed in October, he liked cheesesteaks:)
Yes, he has passed away. At only 69 years of age :(. I considered him the greatest of the two that have been pushing security and privacy from day one, but he needed absolutely no publicity as his work and his staunch uncorruptable commitment to facts did that for him.
Anyway, this paper works out how AIs training AIs leads to what has been termed "model collapse": https://www.lightbluetouchpaper.org/2023/06/06/will-gpt-models-choke-on-their-own-exhaust/.
The only thing that is surprising is that US Government hasn't directed its IT to globally block access to that thing.
The EU Parliament blocked OneDrive from its Office 2010 version on grounds that it didn't want internal documents floating about on Borkzilla's US servers. I would be very surprised if this kind of thing was deemed acceptable by any responsible government anywhere.
I'm guessing banks will be at the forefront of demonstrating just what it is they consider reliable for the security of their customers. Maybe government institutions should align themselves more on that . . .
Personally, I think Copilot, and all supposedly "AI" endeavours should be banned until it is proved that it can be securely controlled. Unfortunately, I reckon this would therefore have to be a perpetual banning order.
Supposed "AI" (LLMs) are all being hyped up by Marketing departments at the moment, and we risk it being taken totally out of Engineering's control. We all know that Marketeers, like Advertisers, cannot be trusted. They are all like kids with new toys at Christmas.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
