Hegseth Announces Grok Access to Classified Pentagon Networks

ThatRobGuy · Thursday at 10:20 AM

7thKeeper said:
Anthropic is out from what it seems. They refuse to remove guardrails from their model that Hegseth is insisting on.

And no matter what version (free/professional) you are using, if it's an LLM, you will never be able to get rid of hallucinations. It's unfortunately baked into how the models function and you can only try to mitigate it, but it will always be in here in LLMs.

Seems like Anthropic is changing their tune pretty quickly

https://edition.cnn.com/2026/02/25/tech/anthropic-safety-policy-change

“Rather than being hard commitments, these are public goals that we will openly grade our progress towards,” the company said in its blog post.

The change comes a day after Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a Friday deadline to roll back the company’s AI safeguards, or risk losing a $200 million Pentagon contract and being put on what is effectively a government blacklist.

With regards to the other aspect you mentioned, while the hallucinations you speak of are certainly pronounced in the free tiered versions, the higher tier versions and enterprise solutions are a different user experience. Where the free tiered ones on the web are basically just a glorified LLM/google search hybrid made for speed, the more robust offerings are more tailored to accuracy and depth.

So while one may be used to hopping on ChatGPT, Perplexity, or Claude and typing a question, and getting an answer back in 5 seconds. The more robust solutions are asking follow-up questions, and giving users the option to prioritize different aspects of what they're trying to do. For example, when utilizing the top tier version of Claude Code, it's not unusual to basically have an 1 hour+ discussion with it, and have it "think" for 5-10 minutes between answers and build a complex solution that can pass SOC2, HIPAA, and PCI audits, where as if you hop on the free version of Claude that's online, it'll spit out a quick and dirty html page with some vanilla JavaScript.

To describe the difference in the past, I've used the steak dinner analogy.

You can order a steak dinner at Cracker Barrel or Applebee's.
vs
You can order a steak dinner from Ruth's Chris or Fleming's

Both will involve sitting a table and having someone bringing you a steak to eat that's edible, but you'll be getting a very different product offering and experience at the latter.

That's not to say that you can't end up with a poorly cooked steak at Ruth's Chris, it obviously happens once in a while. But your chances of getting an overcooked piece of shoe leather, or one that's still cold in the middle (because they didn't thaw properly) when you ordered medium-rare is much less likely at the latter.

MarcusGregor · Thursday at 12:38 PM

AIs can’t stop recommending nuclear strikes in war game simulations

Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

www.newscientist.com

"Would you like to play a game?"

ThatRobGuy · Thursday at 2:37 PM

MarcusGregor said:
Related:

AIs can’t stop recommending nuclear strikes in war game simulations

Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

www.newscientist.com

"Would you like to play a game?"

Some of the article appears to be behind the paywall...

I was able to see as far as this part:
Kenneth Payne at King’s College London set three leading large language models – GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash – against each other in simulated war games. The scenarios involved intense international standoffs, including border disputes, competition for scarce resources and existential threats to regime survival.

Not to keep harping on the "version is everything here folks" talking points but.

Using several-versions old free tier models doesn't necessarily prove what would happen in a realistic application.

For example, they mentioned using the 1 year old free-tier Claude implementation (they're currently on Sonnet 4.6 for the free tier, and Opus 4.6 for premium use)

And there's a big difference between the two

Claude Sonnet 4.6 vs Opus 4.6: Which Model Should You Choose? (2026 Comparison) | NxCode

Claude Sonnet 4.6 vs Opus 4.6 detailed comparison. Benchmarks, pricing, features, and real-world performance compared. Find out which Claude model is right for your use case in 2026.

www.nxcode.io

GPQA Diamond is where the gap becomes dramatic. This benchmark measures PhD-level questions across physics, chemistry, and biology. Opus 4.6's 91.3% vs Sonnet 4.6's 74.1% represents a 17-point chasm — the single largest performance difference between the two models. If your work involves expert-level reasoning, Opus is in a different league.

Similar story with OpenAI. Their "deeper reasoning" is determined based on mode. Meaning, the answers/approaches you'd get from Sonnet 4.6/GPT5.2 are very different than what you'd get with Opus4.6/o1 Pro-Thinking Mode.

Basically, this guy's experiment is tantamount to someone reproducing an already known flaw from a several-versions old, un-patched release of Windows 10, and using that as an argument for why it's dangerous for a government entity to be using the latest & greatest version of Windows Server that's fully patched and up-to-date.

RDKirk · Friday at 9:31 AM

mindlight said:
Musk is a Trump supporter while much of the rest of Silicon Valley are just doing what they need to survive him. It seems that military decisions as to which platform is reliable and which not are being made on the basis of personal loyalty to Emperor Trump.

Personally I found Grok to be infected with Musk's atheism when I asked it questions about Bible dating for example.

The biggest question with this AI as with all of them in fact is to do with the tendency to hallucinate based on statistical models that do not map to reality. I would be interested in what legal constraints and what correspondence to truth tests will be in place for this system.

Also how will command and control hierarchies integrate with the workflows of decision making. AIs can identify targets but it will still need some kind of empirical test and a human decision to fire.

And that leads us to the current conflict between Hegseth and Anthropic, the corporation operating the AI system called "Claude."

Hegseth is threatening to destroy Anthropic because they refuse to develop fully autonomous weapons that take humans out of the loop entirely and automate selecting and engaging targets. I suspect it's because he can't find enough soldiers who will reliably obey illegal orders.

Here is the full statement of Anthropic's CEO on the dispute:

Here is a detailed news story on the dispute.

RDKirk · Friday at 9:36 AM

ThatRobGuy said:
Seems like Anthropic is changing their tune pretty quickly

https://edition.cnn.com/2026/02/25/tech/anthropic-safety-policy-change

“Rather than being hard commitments, these are public goals that we will openly grade our progress towards,” the company said in its blog post.

The change comes a day after Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a Friday deadline to roll back the company’s AI safeguards, or risk losing a $200 million Pentagon contract and being put on what is effectively a government blacklist.

No, that new policy does not indicate a reversal from Anthropic's resistance to creating an LLM that can target humans without human control.

Hans Blaster · Friday at 9:42 AM

RDKirk said:
And that leads us to the current conflict between Hegseth and Anthropic, the corporation operating the AI system called "Claude."

Hegseth is threatening to destroy Anthropic because they refuse to develop fully autonomous weapons that take humans out of the loop entirely and automate selecting and engaging targets.

I normally don't say this, but destroy Anthropic? Don't threaten me with a good time, Sec. Pete.

RDKirk said:
I suspect it's because he can't find enough soldiers who will reliably obey illegal orders.

This is entirely too plausible.

Nithavela · Friday at 10:11 AM

MarcusGregor said:
Related:

AIs can’t stop recommending nuclear strikes in war game simulations

Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

www.newscientist.com

"Would you like to play a game?"

"A strange game. The only winning move is to wipe out mankind."

ThatRobGuy · Friday at 10:23 AM

RDKirk said:
No, that new policy does not indicate a reversal from Anthropic's resistance to creating an LLM that can target humans without human control.

If you read the fine print of the policy change, it does significantly weaken their prior promises.

Or, at the very least, it's a reversal on what their perceived "commitment" level was. They've long tried to brand themselves as the "ethical AI" company.

Anthropic Drops Flagship Safety Pledge

In an abrupt shift, the company may release future AI models without ironclad safety guarantees

time.com

However, when you look at the details of this policy change, it's basically tantamount to saying "Yeah, as long as we're out in front, we'll stick with the commitment to safety, however, if other competitors start to surpass us and start making more money, we'll just match whatever their level of guardrails are"

Aryeh Jay · Friday at 10:24 AM

Nithavela said:
"A strange game. The only winning move is to wipe out mankind."

Where is HAL when you need it.

MarcusGregor · Friday at 10:41 AM

Aryeh Jay said:
Where is HAL when you need it.

"I'm sorry, Dave. I'm afraid I can't do that."

RDKirk · Friday at 10:59 AM

ThatRobGuy said:
If you read the fine print of the policy change, it does significantly weaken their prior promises.

Or, at the very least, it's a reversal on what their perceived "commitment" level was. They've long tried to brand themselves as the "ethical AI" company.

Anthropic Drops Flagship Safety Pledge

In an abrupt shift, the company may release future AI models without ironclad safety guarantees

time.com

However, when you look at the details of this policy change, it's basically tantamount to saying "Yeah, as long as we're out in front, we'll stick with the commitment to safety, however, if other competitors start to surpass us and start making more money, we'll just match whatever their level of guardrails are"

That new policy does not indicate a reversal from Anthropic's resistance to creating an LLM that can target humans without human control.

mindlight · 2026-02-28T10:55:51-0600

RDKirk said:
And that leads us to the current conflict between Hegseth and Anthropic, the corporation operating the AI system called "Claude."

Hegseth is threatening to destroy Anthropic because they refuse to develop fully autonomous weapons that take humans out of the loop entirely and automate selecting and engaging targets. I suspect it's because he can't find enough soldiers who will reliably obey illegal orders.

Here is the full statement of Anthropic's CEO on the dispute:

Here is a detailed news story on the dispute.

It seemed that Hegseth in effect asked anthropic to trust his usage and assume that it would be lawful. So he did not want the company to introduce their own safeguards to the technology suggesting these would be provided by the military in the way they used it. But Anthropic would most probably object that since it is their tech they know what it takes to provide the necessary safeguards and these need to be provided by them and especially when it comes to domestic surveillance and autonomous systems.

RDKirk · 2026-02-28T11:29:44-0600

mindlight said:
It seemed that Hegseth in effect asked anthropic to trust his usage and assume that it would be lawful. So he did not want the company to introduce their own safeguards to the technology suggesting these would be provided by the military in the way they used it. But Anthropic would most probably object that since it is their tech they know what it takes to provide the necessary safeguards and these need to be provided by them and especially when it comes to domestic surveillance and autonomous systems.

I had a chat with ChatGPT (OpenAI) about this a couple of days ago. ChatGPT told me a couple of interesting things:

1. Claude's safeguards are so deeply embedded in the code that removing them would render Claude "unpredictable."

2. ChatGPT has no such safeguards and would be more suitable for Hegseth's apparent purposes.

The second issue in the matter is that even if Anthropic and the DoD can't come to an agreement, they should be able to simply shake hands and walk away. Instead, Hegseth is petulantly threatening to destroy the company by prohibiting any other company that does or would like to do business with the government from doing business with Anthropic.

mindlight · 2026-02-28T12:32:31-0600

RDKirk said:
I had a chat with ChatGPT (OpenAI) about this a couple of days ago. ChatGPT told me a couple of interesting things:

1. Claude's safeguards are so deeply embedded in the code that removing them would render Claude "unpredictable."

2. ChatGPT has no such safeguards and would be more suitable for Hegseth's apparent purposes.

The second issue in the matter is that even if Anthropic and the DoD can't come to an agreement, they should be able to simply shake hands and walk away. Instead, Hegseth is petulantly threatening to destroy the company by prohibiting any other company that does or would like to do business with the government from doing business with Anthropic.

To me that suggests that Claude is an inferior tool in this case. I suspect that investors like Amazon and Google have their own slant on what answers are offensive, suitable or dangerous. I have always preferred the way that I can set my own parameters to questions for the Chat GPT AI and get answers which I, as an expert in some fields, can recognize as authoritative.

In a way your original comment about Hegseth's disdain for Claude is therefore a little disingenuous.

If the tool cannot adapt to the militaries own human assessments and parameter settings then it is trying to impose the very autonomous automatic AI that it is trying to code out of its possible responses. The AI just needs to provide useful answers and be trustworthy in following defined parameterized contextually relevant military protocols not make decisions for the military.

RDKirk · 2026-02-28T14:56:30-0600

mindlight said:
To me that suggests that Claude is an inferior tool in this case. I suspect that investors like Amazon and Google have their own slant on what answers are offensive, suitable or dangerous. I have always preferred the way that I can set my own parameters to questions for the Chat GPT AI and get answers which I, as an expert in some fields, can recognize as authoritative.

In a way your original comment about Hegseth's disdain for Claude is therefore a little disingenuous.

If the tool cannot adapt to the militaries own human assessments and parameter settings then it is trying to impose the very autonomous automatic AI that it is trying to code out of its possible responses. The AI just needs to provide useful answers and be trustworthy in following defined parameterized contextually relevant military protocols not make decisions for the military.

I think you have not been paying attention or are being a bit disingenuous yourself.

Anthropics has been under contract with the DoD since July 2025. In February, 2026, Hegseth moved to change the contract wording. Anthropics claims that the change requires them to remove safeguards that presently prevents their software from accomplishing widespread domestic surveillance and fully automated human targeting.

(I would point out that DoD domestic surveillance has been illegal for 50 or more years. When I took a ride on an RC-135 training mission in the early 80s (electronic intelligence collection for the NSA, which is part of the DoD), we weren't allowed even to listen to US pop music radio stations--the prohibition against DoD domestic surveillance is that tight. The PATRIOT Act loosens it around the periphery--the NSA can listen to a communication that originates from a foreign nation into the US--but widespread domestic surveillance by the DoD is still prohibited by law.)

In my military service, I also spent some time writing scenarios for military exercises. I can't imagine a scenario in which we would want an automaton engaging targets without human supervision. Identifying possible targets is one thing; going through the cycle of destroying them without human supervision is much more something else.

Why would the US government want either of those capabilities, and why would the government not, rather, insist on those safeguards?

mindlight · 2026-02-28T23:39:04-0600

RDKirk said:
I think you have not been paying attention or are being a bit disingenuous yourself.

Anthropics has been under contract with the DoD since July 2025. In February, 2026, Hegseth moved to change the contract wording. Anthropics claims that the change requires them to remove safeguards that presently prevents their software from accomplishing widespread domestic surveillance and fully automated human targeting.

(I would point out that DoD domestic surveillance has been illegal for 50 or more years. When I took a ride on an RC-135 training mission in the early 80s (electronic intelligence collection for the NSA, which is part of the DoD), we weren't allowed even to listen to US pop music radio stations--the prohibition against DoD domestic surveillance is that tight. The PATRIOT Act loosens it around the periphery--the NSA can listen to a communication that originates from a foreign nation into the US--but widespread domestic surveillance by the DoD is still prohibited by law.)

In my military service, I also spent some time writing scenarios for military exercises. I can't imagine a scenario in which we would want an automaton engaging targets without human supervision. Identifying possible targets is one thing; going through the cycle of destroying them without human supervision is much more something else.

Why would the US government want either of those capabilities, and why would the government not, rather, insist on those safeguards?

The USA can define it's own laws and should adhere to what it decides. Where the safeguards that protect that compliance are defined is the issue here. Hegseth says the DOW should define lawful use rather than it's Big Tech supplier.

It is naive to believe that the modern battlefield will not include fifth columnists who are themselves American citizens on US soil.

Regarding autonomous AI killing systems I tend to agree on the macro level. Though the option of setting a "sentinel gun" on a narrow corridor of potential assault by the enemy remains a viable option.

Pommer · 2026-03-01T02:52:03-0600

mindlight said:
It is naive to believe that the modern battlefield will not include fifth columnists who are themselves American citizens on US soil.

Oh goodie, the purity tests are nearly here, who should we “suspect”? I’m rapt.

Nithavela · 2026-03-01T03:07:47-0600

mindlight said:
It seemed that Hegseth in effect asked anthropic to trust his usage and assume that it would be lawful. So he did not want the company to introduce their own safeguards to the technology suggesting these would be provided by the military in the way they used it. But Anthropic would most probably object that since it is their tech they know what it takes to provide the necessary safeguards and these need to be provided by them and especially when it comes to domestic surveillance and autonomous systems.

Just yesterday I saw a social media post of a highly paid AI security expert accidentally having their AI delete their E-Mail folder. I doubt anyone really "knows" AI.

Nithavela · 2026-03-01T03:08:20-0600

Pommer said:
Oh goodie, the purity tests are nearly here, who should we “suspect”?

I suggest implanting every US american with a micro-bomb in their neck, so that they can be retired when AI determines them to be a fifth columnist.

Pommer · 2026-03-01T04:50:32-0600

Nithavela said:
I suggest implanting every US american with a micro-bomb in their neck, so that they can be retired when AI determines them to be a fifth columnist.

That’s down the road, we still have to get through the many rounds of “denunciations” firstly.

We hope the site problems here are now solved, however, if you still have any issues, please start a ticket in Contact Us

Hegseth Announces Grok Access to Classified Pentagon Networks

Part of the IT crowd

New year, new you...

Part of the IT crowd

Alien, Pilgrim, and Sojourner

Alien, Pilgrim, and Sojourner

Call Me Al

you're in charge you can do it just get louis

Part of the IT crowd

Stuck on a ship.

New year, new you...

Alien, Pilgrim, and Sojourner

See in the dark

Alien, Pilgrim, and Sojourner

See in the dark

Alien, Pilgrim, and Sojourner

See in the dark

CoPacEtiC SkEpTic

you're in charge you can do it just get louis

you're in charge you can do it just get louis

CoPacEtiC SkEpTic

Similar threads