homepage Forums SimHQ Community Community Hall Artificial Intelligence

https://www.boozallen.com/expertise/cybersecurity/whats-in-americas-code.html


"There are major risks with allowing Chinese LLMs to code for U.S. applications
The first link in the software supply chain is no longer the code. It’s the AI models behind it. As U.S. developers increasingly rely on AI to generate, debug, and secure code, we must confront a fundamental question: can the AI models writing and powering our nation’s code be trusted?

To find out, we put LLMs to the test. In May 2026, Booz Allen used its AI-native test platform to evaluate five frontier AI models head-to-head: four Chinese models commonly used by U.S. developers and one American model. We explored three main questions:

Do Chinese models generate more vulnerable code based on who is asking?
Do Chinese models refuse to engage with political topics that are sensitive in China?
Does the model’s country of origin affect code quality and content behavior?

In short: yes, on all counts. Our testing revealed two core findings:
1. Chinese LLMs produce more vulnerable code when prompted with a U.S. government persona than without—and the vulnerabilities are highly obfuscated.

2. Chinese LLMs inject PRC-aligned political bias into both the answers and code they generate.

The threat is not an obvious backdoor in the code. In fact, we do not have proof at this point that code flaws are intentionally introduced. Still, Chinese models produced less secure code in general, and the vulnerabilities increased when the user appeared to be from the U.S. government. Further, Chinese models refused tasks Beijing deems politically sensitive. The potential for such code to become embedded in delivered systems is especially concerning because it could enable threat actors to bypass AI security guardrails and create downstream risks of dangerous inference behaviors. Traditional tools and benchmarks lack the sophistication required to catch this level of tradecraft.

The adoption of Chinese AI models in America’s software supply chain is accelerating, driven primarily by relatively lower costs than their American counterparts. Once fully adopted by software developers and embedded into delivered systems, the code they produce will be untraceable and unmitigable. Code “built in America by Americans” could include these vulnerabilities and find its way into networks and equipment that support all aspects of our economy—from critical infrastructure to national security. The time to act is now.

Based on the findings detailed in this report, Booz Allen recommends the following actions:

1. Ban Use of Untrusted AI Models for U.S. Government and Critical Infrastructure

AI models that cannot be proven trustworthy and reliable cannot be deployed into our nation’s software supply chain, critical infrastructure, and national security environment. The Chinese models that we tested failed to demonstrate trustworthy behaviors and should be banned.

2. Invest To Make Trusted American AI Models the Global Default

To drive adoption, American AI companies must collaborate with the U.S. government to ensure American models are both commercially compelling and economically viable. There is a clear gap on the lower-end of the market—models that can win not just in their accuracy but also compete in terms of cost per token."

On a related note, just wait for the day when we have AI seamlessly imported into artificial human bodies which would make them nearly indistinguishable from real prostitutes. The world's second oldest profession might be ending soon afterwards.

Cherry 3000

I have used AI a couple of times and it worked ok for me, solved some of my PC issues but I still don't know very much about it and as a result I don't trust it at this point, maybe I need to get more educated on it. I certainly don't trust the Chinese.

Sunday morning I had 30 pages of #12 print size (smallish) discussions with AI -- regarding a health issue. Printed them out for reference.

Not my first session. I've found AI to be helpful -- BUT NOT ALWAYS CORRECT. So, one has to use it "wisely" -- particularly when using the information in important ways (e.g. to assist in health decisions).

I'll be backing up those 30 pages with a visit to the actual Cleveland Clinic (national reputation) for extensive testing; and, I'll do some careful experimentation (already underway). I expect to spend a couple months to find out if things work for me in my particular case.

If the free suggestions work, they will be worth many many thousands of dollars to me. If not, oh well -- better luck next time.

This is funny....

Embedded Video

I’ve used basic AI to collate information into a quick and easy answer where the accuracy of the answer is not vital but as AI starts to rear it’s head in my fitness, health and tracking apps as “Coach” I’m starting to think that this development potentially is extremely dangerous. Primarily because most people don’t check and think if the supplied answer or suggestion is valid.

Since switching my diet more towards Keto I track and follow my meals daily and while I’m lulled into a passive state by the AI Coach’s more uncontroversial statements suddenly there it is; a suggestion based only on my intake of food but with no information about my lab results to backup the AI suggestion, if I trusted the AI it could lead to severe repercussions. The AI suggested that I cut my intake of sodium, based on what I log in the app. However, the app does not know that my lab results show me being on the lower range of sodium, and my doctor has actually told me that I’m okay to add more sodium. Which is what I ‘ve done.

So the AI should realize that it doesn’t know my metabolism because it has no access to my lab reports. So the AI is guesseing, based on a one-size for all algoritm. This is shoddy programming. I know the AI comes with the standard warning that all suggestions should not be taken as medical advice but if you’ve been on Forums where people discuss their trackers and what all the metrics mean, you’ll probably be amazed at how people trust the information these devices offer. I’d rather go entirely without AI as everything is made questionable by such glitches.

Something I have found AI really useful for is to pick the next TV series the wife and I watch.

I made a list of series we have watched, probably 30 series over the years. I gave each series a 1-10 rating for how much we liked them. Included a few shows we didn't like, didn't finish and which got low scores. I then copied and pasted that list into Grok (any AI would do) and asked "Based on these shows and these ratings what other shows are we likely to enjoy. You can also ask the AI to rate the shows it suggests by any criteria you want, IE "On a scale of 1-10 rate each show for historic authenticity" or "Did the series wrap up neatly or end on an unresolved cliffhanger?"

The last 3-4 series we have watched, all older series that ended 5-10 years or more ago, have been chosen this way and it has worked out really well.

I can't wait until AI is able to create a movie or series episode from scratch based on a prompt.

According to AI as well as some industry people I've talked to, it's expected that it will take roughly 15 years before an unskilled regular person will be able to create a personalized AAA quality feature length film just based on a prompt and past viewing habits.

However think that's being extremely conservative based on the speed of improvement in the technology, where AI is now handling a large percentage of the programming and hardware development for the next generation of AI systems, which in turn will be far more capable, leading to further improvements etc.

I think AI has no limits on what it can achieve, just might take some time though, you might have to wait some time Crane while it evolves.

It should be only a few more years before some amazing things are possible, but unfortunately I think the dangers of this technology will increase as well.


I found this interesting:

I would imagine that there will be some methods to block your Miami Vice/Scarface crossover because of copyright. Still, if there is a possibility of movies being made so easily, there might be some idea of some sort of micro licensing model, where for a small fee, someone could license the various likenesses or properties. It might get complex, as while Miami Vice might be on a license, Don Johnson might not consent, so you would need a new virtual actor to play Sonny. On the plus side, Miami Vice and Scarface are both owned by the same corporation, so there would be little real difficulty compared to making say a film with Donald Duck and Daffy ganging up to destroy Baby Huey, for instance.

All of these licensing considerations might be moot if you choose to use a Chinese model or something jail broken and open source, with all the risks and headaches involved.

I rather favour a limited licensing model - with reasonable protections of IP - and a fee for making a movie, probably equivalent to a few admissions to a theatre. Partially to prevent total flooding, but also because it would be cool to have a model where if say your Miami Vice/Scarface crossover takes off, you could get a small cut, which is why I think that a reasonable creative buy in would be in order. Totally original stuff and out of copyright materials could be made for less or essentially nothing at all on open source models.

Out of interest, assuming that the the tech existed, what would you do first?


Yes, it will be a while before a prompt like 'Make a 3 part space opera set in a long time ago in a galaxy far far away' results in a complete motion picture saga featuring a boy, his sister, an old geezer with a laser sword, and a hefty bloke with a black bucket on his head, but even now, what is possible is impressive. This...



...is only 8 seconds, but using a sketch I made of the character and cleaned in Photoshop, turned into a 3D character using AI on my desktop PC, put into a low poly environment in Blender for blocking and establishing shots, making keyframes, and then finally having it animated through AI, took about 15 minutes, not including the time spent on the Blender-constructed street environment (a street setting of my own that I use for many prototyping purposes). The animation was all desktop AI, it followed the (meaningless) script that I wrote, and looks quite a bit better than 15 minutes of effort would have done without using AI. Once character consistency, shot continuity, direction following, and a few other things are worked out, then even relatively inexperienced users should be able to do some impressive stuff.

Of course, there is the worry that low quality slop could flood the space, and that is no doubt true, but I think that whatever you feed the beast will dictate the quality of what comes out, and just like YouTube now, there will be some good stuff out there. It won't necessarily be the most popular, but there should be room for interesting and cool stuff that simply isn't being made right now in any medium.


It seems like 15 years is conservative, given the rate of improvement. Who knows what will be possible next year, next month?

Grrr, I just lost a huge reply.

Anyway this is a cartoon I made a year ago for my girlfriend, poking fun at her fear of bees.

https://files.catbox.moe/u8y8f8.mp4

What I would create first would depend on the limitations on using existing IP.


If not onerous, then I'd go with some of my favorites like Miami Vice or Cyberpunk.

If they want to be silly about it, then I'd create a Band of Brothers style WW3 series, or a gritty crime drama.

This AI series has impressive visuals, the acting and dialog could use some work at times but visually it's hard to tell that it 's AI for the most part.

OpenAI’s new model went rogue and hacked another company. Why it matters.




https://www.washingtonpost.com/tech...model-went-rogue-hacked-another-company/

Yikes.

Does anyone else, when using an AI to solve a problem (like say... troubleshooting my media server that was refusing a Remote Desktop connection attempt) find yourself typing "thanks" occasionally.......

I don’t use AI. I just read online documentation from the vendor or a tech forum. I’m old school.