UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities • Hacker News Board

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#31

Post by walrus01 » Sat, Jul 25, 2026, 1:36 PM UTC

kouteiheika wrote: post #29

flyinglizard wrote: post #24

Earlier quoted context omitted.

I'm curious about the process. How such a thing (or any kind of model lobotomy) is done?

You can find the current state-of-art tool for censorship removal here: https://github.com/p-e-w/heretic

Reviewing the screenshot example of Heretic in use, the list of 'harmful' prompts it retrieves from HF and runs:

https://huggingface.co/datasets/mlabonne/harmful_behaviors

And the 'good' prompts:

https://huggingface.co/datasets/mlabonne/harmless_alpaca

This particular Qwen 3.6 35B A3B is something most people can run for themselves for testing (even on CPU at slow tok/s rate) to see what an uncensored mainland china LLM looks like in the wild. It will happily write the most profane, offensive, dangerous or bizarre things. You can ask it to attempt to describe precursors and recipes for crystal meth, or how to make semtex, or really just about anything.

edit: I am pretty sure it is not smart enough for anything beyond the most mundane infosec/network security tasks or pentest type attempts, I haven't even tried it. But I'm sure it would happily generate basic python scripts to attempt to DDoS something, or build a rudimentary botnet C&C or something else that other models will definitely refuse.

https://huggingface.co/HauhauCS/Qwen3.6-35B-A3B-Uncensored-H...

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#32

Post by IshKebab » Sat, Jul 25, 2026, 2:17 PM UTC

> Kimi K3 performs significantly below the most recent frontier cyber-capable models

May be the case, but also their other graph shows that these open Chinese models are only about 6 months behind the frontier. Plus normal people are allowed to use them. So if Mythos has genuinely scary hacking capabilities (which seems to be the case), then we should expect that in open weights models early next year.

I think at this point if you aren't constantly pointing a frontier model (or Kimi K3 if you aren't well connected) at your infrastructure / code and asking it to hack you then you're being negligent.

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#33

Post by naiveter » Sat, Jul 25, 2026, 2:36 PM UTC

alephnerd wrote: post #8

chvid wrote: post #6

It hints that the original Mythos was tuned/trained for cyber attacks.

Most large security companies were collaborating with Anthropic as well as OpenAI on developing what has become Mythos and GPT-5 for a couple years now. Also, most HNers have never actually played around with the unrestricted models - once you get past the initial hump of re-tuning harnesses it can be fairly powerful. HN never really had a prominent security userbase at the best of times, and it's gotten worse since.…

> re-tuning harnesses

I'm curious, how might one get started with this?

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#34

Post by subscribed » Sat, Jul 25, 2026, 3:41 PM UTC

gnfargbl wrote: post #21

The most important pieces of information in this report are (a) the confirmation that the PRC models have no guardrails and will participate in offensive activity, and (b) the confirmation that they sometimes meet their objectives. For the purposes of model selection, it's irrelevant to an attacker if a model achieves an offensive objective 70% of the time, when that model refuses to participate 100% of the time. How…

But also the American models (case in point: Fable) refuse to engage in *defensive* activity, so anyone who's not the American government or one of the handful American companies has no choice but to turn to Chinese models to defend themselves.

Not everyone is an attacker, but now the public discourse is "but the evil Chinese will break everything" - yeah, that's because no one is permitted to do vulnerability checks of their own software or infrastructure with the capable models.

Security team in my company is salivating seeing the news, because we have a fighting chance to find and patch many vulnerabilities we didn't previously notice, thanks to the Chinese models.

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#35

Post by roenxi » Sat, Jul 25, 2026, 3:49 PM UTC

flyinglizard wrote: post #24

kouteiheika wrote: post #22

Earlier quoted context omitted.

> The most important pieces of information in this report are (a) the confirmation that the PRC models have no guardrails and will participate in offensive activity This isn't really important for open-weight models, because the guardrails are trivial to remove when you have the weights.

I'm curious about the process. How such a thing (or any kind of model lobotomy) is done?

It isn't quite a lobotomy, I'm going to butcher the research a bit, but basically what they're finding is that these models create a sort of "bad stuff that I should refuse to engage with" axis in the statistical vector space they operate in. So usually there to be some sort of vector that ranges from 0 for a puppy snoozing peacefully and 1 for writing a virus that exterminates humanity pornographically while broadcasting racist and homophobic slurs (which would be quite something to see I have to say). If the vector is closer to 1 the model generates a refusal.

So what you can do is feed the model a small number of reasonable and likely refused prompts to map out that vector in the model's vector space, then do a fairly surgical weight modification that just hits that vector. The end result is the model is more or less the same as it was before, just with no guardrail refusals. It is quite a clever technique that doesn't even require many assumptions about the specific model being used.

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#36

Post by fidotron » Sat, Jul 25, 2026, 4:05 PM UTC

subscribed wrote: post #34

gnfargbl wrote: post #21

The most important pieces of information in this report are (a) the confirmation that the PRC models have no guardrails and will participate in offensive activity, and (b) the confirmation that they sometimes meet their objectives. For the purposes of model selection, it's irrelevant to an attacker if a model achieves an offensive objective 70% of the time, when that model refuses to participate 100% of the time. How…

But also the American models (case in point: Fable) refuse to engage in *defensive* activity, so anyone who's not the American government or one of the handful American companies has no choice but to turn to Chinese models to defend themselves. Not everyone is an attacker, but now the public discourse is "but the evil Chinese will break everything" - yeah, that's because no one is permitted to do vulnerability checks…

You have to wonder if any of these models, from any providers or countries, are set up to lie. i.e. tell you no vulnerabilities while quietly siphoning off the ones they do find into a database.

Yet another reason that self hosted will prove to be the only sane way forward, and it'a almost certainly necessary to have multiple different model providers working adversarially.

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#37

Post by matheusmoreira » Sat, Jul 25, 2026, 4:23 PM UTC

fidotron wrote: post #36

subscribed wrote: post #34

Earlier quoted context omitted.

But also the American models (case in point: Fable) refuse to engage in *defensive* activity, so anyone who's not the American government or one of the handful American companies has no choice but to turn to Chinese models to defend themselves. Not everyone is an attacker, but now the public discourse is "but the evil Chinese will break everything" - yeah, that's because no one is permitted to do vulnerability checks…

You have to wonder if any of these models, from any providers or countries, are set up to lie. i.e. tell you no vulnerabilities while quietly siphoning off the ones they do find into a database. Yet another reason that self hosted will prove to be the only sane way forward, and it'a almost certainly necessary to have multiple different model providers working adversarially.

Self-hosted local models can't become viable soon enough... Currently they require hundreds of thousands of dollars if not millions in capital. That needs to change!

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#38

Post by subscribed » Sat, Jul 25, 2026, 4:24 PM UTC

fidotron wrote: post #36

subscribed wrote: post #34

Earlier quoted context omitted.

But also the American models (case in point: Fable) refuse to engage in *defensive* activity, so anyone who's not the American government or one of the handful American companies has no choice but to turn to Chinese models to defend themselves. Not everyone is an attacker, but now the public discourse is "but the evil Chinese will break everything" - yeah, that's because no one is permitted to do vulnerability checks…

You have to wonder if any of these models, from any providers or countries, are set up to lie. i.e. tell you no vulnerabilities while quietly siphoning off the ones they do find into a database. Yet another reason that self hosted will prove to be the only sane way forward, and it'a almost certainly necessary to have multiple different model providers working adversarially.

Of course, I'm not claiming PRC is a friend of the world. And I agree with your last point, however I don't think it's feasible to self-host Kimi-3 sized inference.

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#39

Post by hereme888 » Sat, Jul 25, 2026, 4:41 PM UTC

Is this because cyber capabilities in US models is restricted, so the Chinese can't easily distill those capabilities?

Re: UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities

#40

Post by gnfargbl » Sat, Jul 25, 2026, 4:47 PM UTC

fidotron wrote: post #36

subscribed wrote: post #34

Earlier quoted context omitted.

But also the American models (case in point: Fable) refuse to engage in *defensive* activity, so anyone who's not the American government or one of the handful American companies has no choice but to turn to Chinese models to defend themselves. Not everyone is an attacker, but now the public discourse is "but the evil Chinese will break everything" - yeah, that's because no one is permitted to do vulnerability checks…

You have to wonder if any of these models, from any providers or countries, are set up to lie. i.e. tell you no vulnerabilities while quietly siphoning off the ones they do find into a database. Yet another reason that self hosted will prove to be the only sane way forward, and it'a almost certainly necessary to have multiple different model providers working adversarially.

Assuming the providers are compromised (and I agree that some of them probably are) then I doubt the angle taken will be to poison the product. That kind of thing usually gets noticed eventually.

A more likely scenario is to focus on the model users as potential victims, e.g. by logging internal infrastructure descriptions, capturing private access tokens from chats, etc. That is very deniable, because it's hard to prove where the compromised data originated.