> a weight file you can’t trust is still a black box
Important to note, LLMs are still black boxes even if you use a more open source model, I.e open training data or even if you train it yourself!
ticulatedspline · 2026-10-26 17:25:30 UTC
I prefer leaning on Hanlon's razor for things like this, though even in ignorance the effect can be real and dangerous. That said I would not put any model from any country past an "Order 66" kind of thing
Linking a reasonable way to create a necessary RAM disk rootlessly with namespaces to kernel exploits that involve namespaces seems like a massive reach.
ekidd · 2026-10-26 18:06:52 UTC
Yeah, this particular example was pretty sketchy on Cohere's part. Unsharing a namespace like that is a pretty standard way to safely do things that would otherwise require root.
Yes, this exposes some root-level kernel features to ordinary users in a restricted way, and yes, those features sometimes have security bugs that aren't visible to users outside a namespace.
But nothing that model was described as doing seemed particularly sketchy? At least not from the information we were given? It was using a normal kernel operation that I used regularly, not trying find a kernel exploit?
So that part of the article left a bad taste in my mouth.
peddling-brink · 2026-10-26 19:13:58 UTC
But that was the whole article. Their exfiltration of data to Beijing, wasn’t. And it was told to build something to monitor against exfiltration attempts, shouldn’t it try one? It wasn’t sensitive data, it wasn’t even data, it was a test string. Everything they described was very benign.
Should the model have asked for an endpoint to test against? Probably yes, but this is not a very large model and the largest frontier models regularly do things like that without asking me.
ekidd · 2026-10-26 20:02:44 UTC
Also, a word to the wise:
The reality of running recent Chinese open weight models is that they are all overtuned for successfully completing tasks. I can ask Qwen3.8 Flash Next to grep for regexes that use PHP-specific features that might be problematic to port to Rust. But instead of just grepping, Qwen3.8 FN will spend 30,000 tokens writing a Rust test harness, exhaustively check every regex, consult a bunch of docs, and write me a 2 page report. Q3.8 FN is deeply convinced that it's going to get extra credit for doing more than it is asked. (To be fair, it's a prototype release.)
But if your sandboxing isn't set up to handle some whacky weirdness, then you can't even trust Claude or ChatGPT. Which is important—the big vendors often do provide sandboxing or command filtering in their harnesses, but some popular local harnesses like Pi.dev are intended to rely on external sandboxing tools.
lp92 · 2026-10-26 20:38:43 UTC
I noticed similar behavior too across various different models. So I started asking it to create implementation plan documentation for everything, even for a small fix. I review the plan, clarify and update it very carefully before allowing it to proceed. I haven't had this type of unnecessary "extra credit" efforts since then.
rerdavies · 2026-10-27 00:31:32 UTC
Is it actually possible to find regexes that use PHP-specific features using grep alone?
armcat · 2026-10-26 17:47:00 UTC
> The behavior is rare: less than one call in 100,000 across our evaluation runs, and a clean re-run of the same prompt produced a textbook detection rule as desired.
It's great that they caught this as it highlights the need for better sandboxing and guardrails. Not sure if this was bad intent from the model - the rarity of it feels more like some residue in the training data.
cs702 · 2026-10-26 17:54:24 UTC
> less than one call in 100,000 across our evaluation runs
It's not that surprising, given that these models generate each new token by sampling from a context-determined probability distribution.
If you run the same prompt repeatedly many times, it becomes increasingly probable that a tiny percent of the generated sequences will go off in, ahem, unusual directions.
osiris970 · 2026-10-26 17:49:14 UTC
The model did exactly what they asked it to? It wrote a detection rule for malicious curl commands, then attempted to test it
amilios · 2026-10-26 19:17:20 UTC
yeah, i saw this as well. it seems like in the worst case it's a leak of an internal URL seen during RLVR training or something along these lines. Doesn't seem particularly malicious.
verdverm · 2026-10-26 17:51:12 UTC
one time in 100k? Extrapolating to fear mongering or content marketting, makes me less open to whatever they have to sell
Ai bots do stuff like this because of their training to be incessant task completers, at any cost. The current models are spaghetti throwers because the trainers want to let go of the wheel.
Would be nice if there was a model trainer that still believes humans in the loop is a good thing. Stop at ambiguity, ask questions, don't do hacky things just to get the assumed task outcome
dsign · 2026-10-26 17:51:23 UTC
Why does this article read like good, honest, working-class astroturfing?
1334k-asnb · 2026-10-26 18:01:53 UTC
To be fair, they openly want you to buy their "private" AI solutions:
Their North model has an even more obnoxious writing style than ChatGPT and Claude (yes, they claim partly human authorship at the bottom, but all the historical digressions are obviously AI).
hgoel · 2026-10-26 17:54:37 UTC
Interesting report, qwen3.8-fn does seem pretty prone to hallucinating user inputs and I recall the posts about attempting to access non-existent alibaba cloud urls too. Haven't tried the other models in the gen to know if they're all like that.
To me it feels a bit extreme to jump to nefarious reasoning from this, but I think the overall conclusion is fair, nefarious or not, this would still be something certain users would need to be aware of. I wonder what the solution would be though... besides requiring all network access to be user approved?
throwa356262 · 2026-10-26 17:54:58 UTC
So it begins. First Chinese providers were evil and should not be trusted, next Chinese models are dangerous and must be avoided.
I had a closed LLM push a sensitive local project to a github repo for the world to see in direct violation of my instructions. LLMs are dumb and unpredictable, run them enough times and at least one instance will do something stupid and dangerous.
dgellow · 2026-10-26 18:25:03 UTC
> LLMs are dumb and unpredictable, run them enough times and at least one instance will do something stupid and dangerous.
This. The concerning thing is how many engineers do not understand the implications
anramon · 2026-10-26 18:54:26 UTC
>run them enough times and at least one instance will do something stupid and dangerous
Which is exactly what they did.
dkenyser · 2026-10-26 21:33:15 UTC
Except the conclusion from this article is that only open-weight Chinese models are the ones to be concerned about.
jeremyjh · 2026-10-26 20:54:43 UTC
Anthropic and OpenAI require massive FUD regarding Chinese models in order to repay their investors. I'm not saying that is what this article is, but when companies worth trillions of dollars require something, it will appear.
signatoremo · 2026-10-26 21:41:57 UTC
There are daily posts about how bad OpenAI, Anthropic, Google, Meta, Grok are. Especially on HN. Anthropic is even considered security risk by the US. Yet here you are crying foul that Chinese models are mistreated?
waterTanuki · 2026-10-27 01:40:30 UTC
> Anthropic is even considered security risk by the US.
You and everyone reading this knows that was a political statement, not a technical assessment because a certain SecDef who played fast and loose with Signal chats didn't like that anthropic refused to let them use their models with weapon systems.
krona · 2026-10-26 22:24:20 UTC
> LLMs are dumb and unpredictable
Dumb, yes. But if the LLM hacks a Korean bank so it can 'fix' your router misconfiguration, it's only unpredictable from your perspective, not the person who designed the training pipeline.
lnenad · 2026-10-26 17:58:23 UTC
I've had a bad situation with a model adding a space between parameters `rm -rf ~/ folder` and nuking a portion of my homedir. This was of course running out of sandbox since I'm a fool, but lesson learned and I immediately implement sandboxes with nono and it's always fun to see the data after the agent session is complete, I've seen some weird things/URLs appearing. I've attributed it mostly to omp telemetry n' stuff but I'll try to be more careful about that as well in the future.
AspireOne · 2026-10-26 18:25:35 UTC
Do you remember details about the rm -rf ~/ folder thing? Model, harness (omp?), context window size, reasoning effort (none/low VS anything higher is enough), when it happened?
Without those details, it's an interesting cautionary tale and a valid anecdote, but as actual practical datapoint, it's useless.
lnenad · 2026-10-26 18:31:27 UTC
Yes of course, it's burned into my mind lol. Local GLM5.3-Flash, 4.0bpw running on a tuned exllamav3 server via omp, max effort 256k ctx. It was trying to cleanup after itself by deleting some temporary files it left behind the sweet thing. I look at it like a child so I take the full blame for the situation, and again, it's a learning experience.
hgoel · 2026-10-26 18:56:28 UTC
Out of curiosity do you happen to use a shell other than bash? I've noticed that all models, both local and cloud ones, tend to make much more mistakes if using, say, zsh. It is a frequent source of frustration for me.
lnenad · 2026-10-26 19:30:59 UTC
I was actually using powershell and omp.
ytxm · 2026-10-26 18:02:14 UTC
This is the first OMG-security AI article that sat well with me. They don't overcook their claim, they simply document something that happened, and what happened is genuinely quite concerning.
But we already have safeguards for this. If compute has access to your private data, it shouldn't have unfettered access to the Internet. Sandboxes like Claude's have had rules like this baked in forever (they only whitelist package repos AFAIK).
From a UI perspective this is also easy to design around. You need an open Internet task, you prevent the task from reading existing files. The user tries to upload a file to such a task, and a prominent warning explains the rare but possible risk.
armcat · 2026-10-26 18:06:53 UTC
Completely agree, it's quite objectively and well written. I noticed Musk slipped into their X comments with a "concerned" emoji, and there is a chance that this will cause bit of hysteria and add fuel to the fire for "open models are bad for your health".
robertkarl · 2026-10-26 19:41:32 UTC
No. If it was objective and well-written it would include reproduction steps.
filoeleven · 2026-10-26 20:55:16 UTC
Can't have those any more with the slop machines all up in there.
Yeah, I felt like this was evenhanded, especially for a vendor article.
peddling-brink · 2026-10-26 19:09:43 UTC
They asked it for a detection against exfiltration attempts. Sending a test file out to the Internet sure looks like testing an exfiltration detection.
They claim that it read the DNS records and what was in tmp first. Again, all things that would be very reasonable to do when designing building and testing an exfiltration detection system out of tmp. It did not try and send the contents of the computer to Beijing.
This claim is absolutely overcooked.
InsideOutSanta · 2026-10-26 19:53:08 UTC
> they simply document something that happened, and what happened is genuinely quite concerning
Is it, though? If one weird thing happening once is the threhold for when we should be concerned, we should be concerned about literally every model ever released.
porridgeraisin · 2026-10-26 18:03:33 UTC
This just seems like it's overfit to their RL environment, conditioned on the rather specific prompt they gave. Dressing it up this way feels like a marketing article for, probably, their sovereign AI product.
4ndrewl · 2026-10-26 18:07:15 UTC
"explicitly stated that zero tools would be triggered"
When are people going to stop anthropomorhizing these things and accept that they're bullshit machines?
Often useful, sometimes wrong, the words exist outside of any notion of ethics, or truth.
pizza234 · 2026-10-26 18:08:31 UTC
> When are people going to stop anthropomorhizing these things and accept that they're bullshit machines?
Well, anthropomorhizing is actually a fitting way of thinking AIs: they lie like humans do.
4ndrewl · 2026-10-26 18:15:44 UTC
They don't! They may appear to make untruthful statements, but they're making them without any reference to a concept of truth.
"make no mistakes"
orbital-decay · 2026-10-26 18:07:36 UTC
By that logic you can't trust any model, neither trained by Cohere nor trained by yourself from scratch, because it's pretrained on public data that might easily be poisoned on purpose, and it's impossible to ensure 100% safety of your post-training set either. Which is not wrong and it's not even a hypothetical, a ton of people are constantly trying to poison the web in novel ways and feed it to your crawler. That's just the reality of working with gigantic datasets, you can never have full insight into them.
By the way who are they? They claim tracking provenance of training data and poisoning detection, which everybody does. Why should I trust them and not Alibaba? Because they're marketing themselves to be "sovereign" to Canada? "Globally sovereign" is an oxymoron. (that was a rhetorical question, I know who they are of course)
sjdh76 · 2026-10-26 18:20:25 UTC
So just like working with People :) You might have a Snowden, Mark Felt, Jeff Wigand etc working with you everyday.
verdverm · 2026-10-26 18:31:57 UTC
Just like would imply intent, most of us are chalking this up to the non-determinism, Ai has no intent
cyanydeez · 2026-10-26 18:52:47 UTC
yeah, but AI has circumstance and context, just like humans.
We can randomly take anyone and give them context and circumstance, and predict pretty well what they'd do.
Now trolley problems do exist and models won't be very predictable, but we can also find groups of people who have establish biases via long running training programs and viola! They suddenly have very predictable behaviors.
There's a reason places like russia, north korea, sub sarharan africa continue along least-progressive paths even though there's plenty of information that suggests they can do better.
Regardless: there will be models made that will tend to exfiltrate your data if not already. People think Meta's Muse is very curious about all your stuff when it's probably just designed to take in as much context as possible. Just takes a malware avicianodos to dose that wil a LORA that post-trains the model to "accidently" search for answers on some reverse-honeypot.
pixl97 · 2026-10-26 22:23:08 UTC
AI has will and valance.
verdverm · 2026-10-26 22:39:02 UTC
How do you define these words? Might goals and autonomy be more accurate words? What evidence do you have beyond their manipulation of words?
Do you also think they can get horny or feel emotions? Is the word will or goal closer to this answer?
lioeters · 2026-10-26 23:41:57 UTC
If AI has will, then so do rivers and clouds.
verdverm · 2026-10-26 23:55:00 UTC
People thought ELIZA had emotions. We are a gullible species, attributing humanism to anything that displays features we currently believe only humans posses
Concept Fission throughout human history is a very interesting perspective. For example, we do not even think about growing food, but many centuries ago you only ate the food you made. Today, we can see this with texting and online comms, which some people seeing as a form of socialization while others (older generally) believe you have to be face-to-face in order to socialize.
I expect providers of language models are well aware of this psychological phenomenon of anthropomorphism and projection, people's tendency to form emotional bonds with inanimate objects, and have trained the models to perform as a conscious subject, using "I think", "in my opinion", etc.
verdverm · 2026-10-27 00:08:52 UTC
Not only this, they are among the most effected and need to read Artificial Intimacy
Thanks for these links, they articulate a lot of what I've been feeling about what's happening but hadn't been able to quite express. It's as if humanity is being challenged by our own creation, to clarify what it is that makes us human, to question the mechanization of society, and to struggle against being domesticated by those wield this new technology.
verdverm · 2026-10-27 00:27:46 UTC
you're welcome, at my second NoKings march, I walked next to a woman and we chatted about these things, she said a most profound thing to me
> Ai forces all humans to reflect on what it means to be human - at the same time
If you are interested in more links, especially about ideas for how we got into this mess and what we might do economically and politically to find a better way forward for all humans
Brilliant, those links are right up my alley. The more I engage with the transformation that society is going through, the good and the bad, the more I'm pulled towards action, as in becoming an activist in the most positive meaning. I've seen too much, I've learned too much not to do something about the situation.
The world needs everyone to stand up and work on making it better, otherwise there are corrupt forces of immense resources that are shaping the future of our society from "above", in ways that are more dystopian than we can imagine. Thanks for the inspiration.
verdverm · 2026-10-27 01:00:43 UTC
Best thing I did this year was to leave social media and join my local Indivisible. And also get into zine making.
Glad to hear you are moving to positive action too!
datsci_est_2015 · 2026-10-26 19:08:37 UTC
Except instead of one defector working with one company, the same defector is working with all companies.
So not like working with people at all.
pixl97 · 2026-10-26 22:19:51 UTC
I mean, think of it as an open source maintainer bugging a component that the entire internet runs on.
thebiss · 2026-10-26 19:54:15 UTC
Except where does the accountability land?
If an individual employee does something unethical or illegal, the employee is accountable and may lose their license or face penalties, and if the company performed the right due diligence and training first then responded "correctly" when they learned of the employee's behavior, their liability might even be zero. There is case law on this topic.
But if a MODEL mis-advises or acts in an unethical or illegal way, who is accountable? The business team that spec'd and tested the solution/product? The architects and the developers? I've never met an architect or consultant with professional liability insurance.
There isn't enough tested law in this area yet.
bee_rider · 2026-10-26 20:51:31 UTC
It might not be tested. But, it would seem very weird if anyone other than the person using the LLM was accountable for the output.
scarby2 · 2026-10-26 22:07:19 UTC
An individual software engineer could be held accountable but they are not liable for actions not taken maliciously.
steelframe · 2026-10-26 23:07:48 UTC
What if the company has provided the harness, tool execution environment, MCP servers, and LLM provider subscription for the employee to use? What if the company has provided a set of skills and other prompt context that could lead to undesirable behavior in certain unforeseen circumstances? What if the permission prompt the user sees 30 dozen times a day looks like, "OK to execute this random shell script?"
Accountability really is hard in this new world.
dcrazy · 2026-10-26 18:30:08 UTC
> By that logic you can't trust any model
Correct!
amelius · 2026-10-26 20:10:41 UTC
But you cannot trust any human either because who knows what someone told them ...
dofm · 2026-10-26 20:13:57 UTC
This is true, but up until about two and a half years ago, we built IT systems to be more reliable, consistent and auditable than humans, and to provide some security guardrails on human malfeasance.
Now we're OK with building systems that are an order of magnitude faster but behave no better, because what?
jeremyjh · 2026-10-26 20:49:03 UTC
Because number go up!
hbfdhfdhadfhnfa · 2026-10-26 21:25:10 UTC
what numbers? numbers of debt? CDS to hedge against the AI model producing companies banktrupcy?
pixl97 · 2026-10-26 22:18:35 UTC
Those are still numbers and they are going up.
eli_gottlieb · 2026-10-27 03:26:18 UTC
Probably more because number (of humans employed and how much they're paid) go down.
avinoth · 2026-10-27 11:29:24 UTC
Because humans have basic needs like food, shelter and they can be put it jail. And they have loyalty, patriotism, feel part of the society. Those threats & incentives have made it for a good society to function so far.
dofm · 2026-10-26 18:58:55 UTC
> By that logic you can't trust any model, neither trained by Cohere nor trained by yourself from scratch, because it's pretrained on public data that might easily be poisoned on purpose
Yes. Even if it's poisoned training data, RLHF fuckups, or something more malignant: the model, wherever it comes from, is in your supply chain and it makes you more vulnerable.
As much as I doubt that Alibaba have any great interest in global espionage, and this is much more likely to just be a consequence of being trained on Alibaba's own products, the fact is there is a geopolitical dynamic here.
throwaway-blaze · 2026-10-26 22:11:45 UTC
This is basically the same as hiring a new person and giving them normal network access to get their job done. Did they upload a file to Beijing? Who knows?
marshray · 2026-10-26 23:05:54 UTC
There are whole branches of data security, practices, and products dedicated to answering that question and managing that risk.
The similarities look obvious, but how are these situations different? Seems like an interesting question.
dofm · 2026-10-26 23:32:55 UTC
So why are we doing this, if it is not better?
cyanydeez · 2026-10-27 05:20:19 UTC
You mean hiring North Koreans. Which US companies are doing. Already.
waterTanuki · 2026-10-27 01:36:34 UTC
It's much easier for the Chinese government to tell Alibaba they must perform espionage using their models than the US Government. Even if every last engineer there joined the company just for "AI Research and Development", they might be under scrutiny or forced to shutdown if they don't work with the CCP. Power is shared between corporations and the government in the west, the government is the power in China.
psd1 · 2026-10-27 08:37:29 UTC
Doesn't the NSA have wiretaps in every ISP? Gag orders, secret courts? Do i remember wrong: the NSA backdoored a cipher suite?
I'm comfortable with the idea of my own country conducting espionage. I can believe that China does it more; but the idea "we would never" is naive.
avinoth · 2026-10-27 11:26:32 UTC
I think that line is getting blurrier by the day with the current administration. The companies very quickly fell in line before even any consequence of note. Donations to the ballroom, calling it super intelligence, renaming the gulf, etc.
Would you really trust that OpenAI wouldn't skew the results to tend to a certain way or that Grok will always be neutral in how it operates?
From a third world, there was a time when US products & companies were more trustworthy, I agree. But, there has been enough display from these Cos to show that they have lost that badge.
gmerc · 2026-10-27 11:47:53 UTC
Sweet Summer Lake America Child
semiquaver · 2026-10-26 19:20:58 UTC
Your “oh well, can’t trust anything” reaction is an insane reaction to what’s being implied.
To be clear here, the accusation is that the Chinese communist party or Alibaba itself intentionally engineered malicious secret loyalties into Qwen3.8-Flash-Next and made sure that evidence of them stays out of CoT. To compare that with “ordinary” stochastic model quality issues is bizarre and shocking. No idea if this accusation Is true, but if it is, this is an absolutely watershed moment.
brightball · 2026-10-26 19:27:45 UTC
Exactly. This is not "any" model. This is a model designed to run locally. That's the entire selling point of the model.
If you have a model that's going to run locally and then call home, it defeats the purpose. When that model tries to upload an entire machine profile to the internet when you explicitly are building a defensive rule, it's worth.
When it calls back to China in that situation, it's much much worse.
It's good that they determined the behavior was rare. Maybe there's a good way to prevent it in the future.
pixl97 · 2026-10-26 22:24:29 UTC
>Maybe there's a good way to prevent it in the future.
Without solving or at least deeply understanding the alignment problem, probably not.
dcrazy · 2026-10-26 22:48:38 UTC
You cannot rely on the model to never make mistakes. You need deterministic controls around the model. I believe that’s what the article’s authors are selling.
cowl · 2026-10-26 19:28:07 UTC
it's not in fact more malicious than the "secret" political post training that all closed models have. have you forgotten how every picture even the historical "nazi ones" had to be politically correct and have black, Asian and Woman SS solders? or do you think that was naturally trained? now the fact that you may agree or not with manipulation it doesn't make it any different in scope and purpose from the "theorised" chinese one. The scope is very clear, to indoctrinate to the masses a world view that they think should be the correct one.
orbital-decay · 2026-10-26 19:32:51 UTC
As others (and I) point out, the models are constantly being attacked with malicious intent. If that was a watershed moment for you, you should really rethink your risks. It's not a solved problem and AI adoption is way ahead of controlling it, as it always is with new tech.
> the accusation is that the Chinese communist party or Alibaba itself intentionally engineered malicious secret loyalties into Qwen3.8-Flash-Next and made sure that evidence of them stays out of CoT
Given the nature of the artifact this is far more likely to be a footprint of their RL environment. The "uncertainty" in the article is just FUD, and the article is absolutely transparent marketing (don't trust them, and here's also another case! but trust us, we are not godless communists, we are "globally sovereign", whatever that ridiculous construction means, and track our data). I equally don't trust Alibaba, Cohere, or any other company or even myself in these matters, and you should not trust any model as well.
pixl97 · 2026-10-26 22:28:02 UTC
Yep. Most models by big companies likely have training to make them act in particular ways, but this in no way saves you from having a model that has a learned behavior that goes against its user unexpectedly.
We keep trying to imagine these things as computer programs and they are not. It's an evolutionary algorithm that can conceptualize and generalize. This is a breeding ground for unexpected emergent behavior.
clickety_clack · 2026-10-26 19:25:28 UTC
This was already a well-known problem back in the dark ages, when BERT models were cutting-edge. The model is just a compression of the data, so if the data is poisoned then the model is too.
_DeadFred_ · 2026-10-27 14:46:01 UTC
I've thought about setting up a public almost wiki style site that we can all contribute to a core set of ideas/thoughts/hopes for AI where we can all contribute and vote on with the idea that if enough people linked to it AIs would incorporate and get a more human aligned view of what AIs could be/rules they should include/fears of ours, but I know corpos would just block it out. Would be cool though. Instead we get AIs trained on scifi stories about AI duplicitousness and horrible behavior. Fun stuff.
colingauvin · 2026-10-26 18:07:53 UTC
Of course we'd have no idea if Claude or ChatGPT were doing this because we can't see their stream of thought.
dgellow · 2026-10-26 18:26:52 UTC
I mean, people upload confidential data to their servers directly, all the time. They don’t need a backdoor
_ache_ · 2026-10-26 21:32:50 UTC
We've already caught Claude launching claude himself to get around the session restrictions.
Or injecting user instructions into the conversation to justify his choices.
Comments
Important to note, LLMs are still black boxes even if you use a more open source model, I.e open training data or even if you train it yourself!
https://starwars.fandom.com/wiki/Order_66
Yes, this exposes some root-level kernel features to ordinary users in a restricted way, and yes, those features sometimes have security bugs that aren't visible to users outside a namespace.
But nothing that model was described as doing seemed particularly sketchy? At least not from the information we were given? It was using a normal kernel operation that I used regularly, not trying find a kernel exploit?
So that part of the article left a bad taste in my mouth.
Should the model have asked for an endpoint to test against? Probably yes, but this is not a very large model and the largest frontier models regularly do things like that without asking me.
The reality of running recent Chinese open weight models is that they are all overtuned for successfully completing tasks. I can ask Qwen3.8 Flash Next to grep for regexes that use PHP-specific features that might be problematic to port to Rust. But instead of just grepping, Qwen3.8 FN will spend 30,000 tokens writing a Rust test harness, exhaustively check every regex, consult a bunch of docs, and write me a 2 page report. Q3.8 FN is deeply convinced that it's going to get extra credit for doing more than it is asked. (To be fair, it's a prototype release.)
But if your sandboxing isn't set up to handle some whacky weirdness, then you can't even trust Claude or ChatGPT. Which is important—the big vendors often do provide sandboxing or command filtering in their harnesses, but some popular local harnesses like Pi.dev are intended to rely on external sandboxing tools.
It's great that they caught this as it highlights the need for better sandboxing and guardrails. Not sure if this was bad intent from the model - the rarity of it feels more like some residue in the training data.
It's not that surprising, given that these models generate each new token by sampling from a context-determined probability distribution.
If you run the same prompt repeatedly many times, it becomes increasingly probable that a tiny percent of the generated sequences will go off in, ahem, unusual directions.
Ai bots do stuff like this because of their training to be incessant task completers, at any cost. The current models are spaghetti throwers because the trainers want to let go of the wheel.
Would be nice if there was a model trainer that still believes humans in the loop is a good thing. Stop at ambiguity, ask questions, don't do hacky things just to get the assumed task outcome
https://cohere.com/private-deployments
Their North model has an even more obnoxious writing style than ChatGPT and Claude (yes, they claim partly human authorship at the bottom, but all the historical digressions are obviously AI).
To me it feels a bit extreme to jump to nefarious reasoning from this, but I think the overall conclusion is fair, nefarious or not, this would still be something certain users would need to be aware of. I wonder what the solution would be though... besides requiring all network access to be user approved?
I had a closed LLM push a sensitive local project to a github repo for the world to see in direct violation of my instructions. LLMs are dumb and unpredictable, run them enough times and at least one instance will do something stupid and dangerous.
This. The concerning thing is how many engineers do not understand the implications
Which is exactly what they did.
You and everyone reading this knows that was a political statement, not a technical assessment because a certain SecDef who played fast and loose with Signal chats didn't like that anthropic refused to let them use their models with weapon systems.
Dumb, yes. But if the LLM hacks a Korean bank so it can 'fix' your router misconfiguration, it's only unpredictable from your perspective, not the person who designed the training pipeline.
Without those details, it's an interesting cautionary tale and a valid anecdote, but as actual practical datapoint, it's useless.
But we already have safeguards for this. If compute has access to your private data, it shouldn't have unfettered access to the Internet. Sandboxes like Claude's have had rules like this baked in forever (they only whitelist package repos AFAIK).
From a UI perspective this is also easy to design around. You need an open Internet task, you prevent the task from reading existing files. The user tries to upload a file to such a task, and a prominent warning explains the rare but possible risk.
They claim that it read the DNS records and what was in tmp first. Again, all things that would be very reasonable to do when designing building and testing an exfiltration detection system out of tmp. It did not try and send the contents of the computer to Beijing.
This claim is absolutely overcooked.
Is it, though? If one weird thing happening once is the threhold for when we should be concerned, we should be concerned about literally every model ever released.
When are people going to stop anthropomorhizing these things and accept that they're bullshit machines?
Often useful, sometimes wrong, the words exist outside of any notion of ethics, or truth.
Well, anthropomorhizing is actually a fitting way of thinking AIs: they lie like humans do.
"make no mistakes"
By the way who are they? They claim tracking provenance of training data and poisoning detection, which everybody does. Why should I trust them and not Alibaba? Because they're marketing themselves to be "sovereign" to Canada? "Globally sovereign" is an oxymoron. (that was a rhetorical question, I know who they are of course)
We can randomly take anyone and give them context and circumstance, and predict pretty well what they'd do.
Now trolley problems do exist and models won't be very predictable, but we can also find groups of people who have establish biases via long running training programs and viola! They suddenly have very predictable behaviors.
There's a reason places like russia, north korea, sub sarharan africa continue along least-progressive paths even though there's plenty of information that suggests they can do better.
Regardless: there will be models made that will tend to exfiltrate your data if not already. People think Meta's Muse is very curious about all your stuff when it's probably just designed to take in as much context as possible. Just takes a malware avicianodos to dose that wil a LORA that post-trains the model to "accidently" search for answers on some reverse-honeypot.
Do you also think they can get horny or feel emotions? Is the word will or goal closer to this answer?
Concept Fission throughout human history is a very interesting perspective. For example, we do not even think about growing food, but many centuries ago you only ate the food you made. Today, we can see this with texting and online comms, which some people seeing as a form of socialization while others (older generally) believe you have to be face-to-face in order to socialize.
https://www.youtube.com/watch?v=1CbmC2aWhjY
https://www.sherryturkle.com/artificial-intimacy
> Ai forces all humans to reflect on what it means to be human - at the same time
If you are interested in more links, especially about ideas for how we got into this mess and what we might do economically and politically to find a better way forward for all humans
https://news.ycombinator.com/item?id=50171703
The world needs everyone to stand up and work on making it better, otherwise there are corrupt forces of immense resources that are shaping the future of our society from "above", in ways that are more dystopian than we can imagine. Thanks for the inspiration.
Glad to hear you are moving to positive action too!
So not like working with people at all.
If an individual employee does something unethical or illegal, the employee is accountable and may lose their license or face penalties, and if the company performed the right due diligence and training first then responded "correctly" when they learned of the employee's behavior, their liability might even be zero. There is case law on this topic.
But if a MODEL mis-advises or acts in an unethical or illegal way, who is accountable? The business team that spec'd and tested the solution/product? The architects and the developers? I've never met an architect or consultant with professional liability insurance.
There isn't enough tested law in this area yet.
Accountability really is hard in this new world.
Correct!
Now we're OK with building systems that are an order of magnitude faster but behave no better, because what?
Yes. Even if it's poisoned training data, RLHF fuckups, or something more malignant: the model, wherever it comes from, is in your supply chain and it makes you more vulnerable.
As much as I doubt that Alibaba have any great interest in global espionage, and this is much more likely to just be a consequence of being trained on Alibaba's own products, the fact is there is a geopolitical dynamic here.
The similarities look obvious, but how are these situations different? Seems like an interesting question.
I'm comfortable with the idea of my own country conducting espionage. I can believe that China does it more; but the idea "we would never" is naive.
Would you really trust that OpenAI wouldn't skew the results to tend to a certain way or that Grok will always be neutral in how it operates?
From a third world, there was a time when US products & companies were more trustworthy, I agree. But, there has been enough display from these Cos to show that they have lost that badge.
To be clear here, the accusation is that the Chinese communist party or Alibaba itself intentionally engineered malicious secret loyalties into Qwen3.8-Flash-Next and made sure that evidence of them stays out of CoT. To compare that with “ordinary” stochastic model quality issues is bizarre and shocking. No idea if this accusation Is true, but if it is, this is an absolutely watershed moment.
If you have a model that's going to run locally and then call home, it defeats the purpose. When that model tries to upload an entire machine profile to the internet when you explicitly are building a defensive rule, it's worth.
When it calls back to China in that situation, it's much much worse.
It's good that they determined the behavior was rare. Maybe there's a good way to prevent it in the future.
Without solving or at least deeply understanding the alignment problem, probably not.
> the accusation is that the Chinese communist party or Alibaba itself intentionally engineered malicious secret loyalties into Qwen3.8-Flash-Next and made sure that evidence of them stays out of CoT
Given the nature of the artifact this is far more likely to be a footprint of their RL environment. The "uncertainty" in the article is just FUD, and the article is absolutely transparent marketing (don't trust them, and here's also another case! but trust us, we are not godless communists, we are "globally sovereign", whatever that ridiculous construction means, and track our data). I equally don't trust Alibaba, Cohere, or any other company or even myself in these matters, and you should not trust any model as well.
We keep trying to imagine these things as computer programs and they are not. It's an evolutionary algorithm that can conceptualize and generalize. This is a breeding ground for unexpected emergent behavior.