Musk’s AI tool Grok will be integrated into Pentagon networks, Hegseth says
Jesus Christ 😲
https://www.theguardian.com/technology/2026/jan/13/elon-musk-grok-hegseth-military-pentagonOpen linkView original on reddthat.comSyndicated from the fediverse. Read and engage on the original instance.
View original on reddthat.comJesus Christ 😲
https://www.theguardian.com/technology/2026/jan/13/elon-musk-grok-hegseth-military-pentagonOpen linkView original on reddthat.com
23 replies
It won't be "integrated" beyond employees using it to summarize emails.
Elon Musk is once again supporting Trump and allies for the elections, so automatically they needed some way to give him money. This contract is one of the ways.
The tool itself is useless.
It could probably put Kegsbreath in a bikini if you asked it to. But yeah, other than that it's useless.
Yep likely/hopefully it will be limited to that. But a) It's Trump and Musk so we can't trust it'll stay at that and b) even emails have a fuckton of restricted info in and I don't trust Musk with it.
They won't be summarizing emails. They'll be mass producing CSAM to use against their enemies and to get themselves off.
Therefore if a single email login is compromised, all info transacted via email is exfiltrated. Not just email from the compromised login. All of it.
Department of Pedophile Defense
Department of Pedophile
DefenseOffenseCan't wait to try asking grok to post about ongoing operations and reading it on xitter
LMAO! It’s going to try and rape the capitol building.
That'll be it's fan fiction... The real danger is letting an "antiwoke" Ai have inside knowledge on state secrets. It's not like that Ai has shown homicidal tendencies to participate groups of people let alone a disgust of humanity as a whole?
When I say Ai (LLM) learning about the atrocities our government doesn't let see the light of day is a mistake. And these idiots don't want to stop there. The want the machines to have the ability to kill without human involvement. They are testing it now in Gaza.
Well that doesn't sound risky at all. :|
All that classified data in Musk's grubby little hands.
He got most of it from DOGE already anyway
That was just everyone's social security information. This is military stuff. Much more useful to his Russian and Saudi pals.
The next president should use this as pretext for the state to shut down Grok for national security reasons.
Ironically I bet the considered this angle when they got together with Musk for this.
“If he gets out of line again, we can seize his most valuable assets since they are not a national security risk.”
They so jealous of the War thunder forums
Wasn't this the cause of future events in the whole Terminator series?
No, in the terminator they had an AI that could think.
Don't need to be able to think to cause mass devastation with a rogue AI. I present the paperclip problem: https://en.wikipedia.org/wiki/Instrumental_convergence
And from my experience trying to use it for programming, I know it absolutely will try its best to reach an unbounded goal you set for it even if it means creating the most convoluted and unmaintainable spaghetti code you've ever seen, or even trying to find loopholes to suppress error messages without actually fixing the error.
I didn't realize the paperclip maximizer was a real thought experiment outside this game I found a few years back: https://www.decisionproblem.com/paperclips/
https://archive.ph/IEPh7
what could go wrong
I work on IT. If I believed in God, then God help us all.