No, you paid for the end result. The thought process is a step in between, a function. Think about it, who should get charged if the answer you received comes from a cache? Thinking tokens are the complexity-of-the-problem cost. I mean, you may not agree but both are valid points of view.
anigbrowl 10 minutes ago [-]
No I didn't. I buy my tokens from a provider that exposes the model reasoning so I can understand what it's doing and work with it, or interrupt if I see things going in the wrong direction.
paxys 57 minutes ago [-]
The only person calling it stealing is the author of this article, so this is a pointless discussion. The majority of this thread is just arguing with themselves.
1 hours ago [-]
__MatrixMan__ 1 hours ago [-]
Liberating!
Groxx 6 hours ago [-]
>We take a trace produced by a frontier model, replay it into a weaker sibling, jailbreak the weaker model, ...
I'm honestly rather curious if this was intentionally allowed, it's the sort of validation that's easy to miss (particularly if you're wading into the vibe waters). Seems like something that'd be absolutely riddled with possibilities for shenanigans.
matthewdgreen 3 hours ago [-]
Author of the blog post here: just for the record, I did try replaying against other models. I replayed from GPT 5.5 to 5.5-mini. It "worked" (replay not rejected) but mini didn't hand me the plaintext of the reasoning. I'm not a jailbreaking expert, so I didn't really "push" on it very hard. I also didn't pursue Opus/Haiku for reasons I don't really recall. These folks took it all the way. I'm bummed it didn't work for me, but having these folks prove the channel out is a very good consolation prize.
If you didn’t allow it, you wouldn’t be able to change models in the same conversation, as key parts of the context would be lost.
Wouldn’t surprise me if the providers just remove that ability and lock the model once the conversation starts.
Groxx 6 hours ago [-]
Fair (I haven't been using the encrypted-reasoning systems, though this is common in open ones - I'm kinda surprised it's an option in encrypted ones too), though what they're doing here is cross-user replays in addition to cross-model.
myworkaccount2 6 hours ago [-]
There seems to be an obvious choice to make here, should you give the users to decrypt and use the COT that they did not generate themselves?
This is only required if you want users to be able to share things with everyone and you are going for the simplest implementation.
If not you could try to keep a record of keys associated with a user, then when a new request comes in look through to see if the user has a valid key to decrypt the COT.
For explicit shares, just add the key used in that one conversation to the users valid keys. For global shares use the global keys. But that's adding more complexity to the system.
supriyo-biswas 4 hours ago [-]
It’s about being able to change models mid-task. For example, I want to be able to plan using Fable but implement the plan using Sonnet, and that won’t work if this is implemented.
SaltyBackendGuy 4 hours ago [-]
Or even my fable credits run out mid task and need to switch back to opus >.<
QuercusMax 3 hours ago [-]
Prior to LLMs I never considered that I might have to make a resource-usage decision between hiring Star Trek's Data vs. his stupider brother B4...
For plan it's relatively easy, just make the plan the artifact. The point is to ingest knowledge with one model and use it in another, and that is not necessarily easily expressible in natural language.
Der_Einzige 6 hours ago [-]
100% guaranteed that this research just forced this to happen now.
Sucks.
dannyw 4 hours ago [-]
It’s already patched according to the authors. Details were not specified.
pas 5 hours ago [-]
or add some metadata and don't allow downgrading.
aix1 6 hours ago [-]
I really don't understand why server-side storage of the trace isn't a viable approach here, with only a unique key flowing to the client and back. Does it have something to do with how backend load-balancing works?
amluto 5 hours ago [-]
Makes no difference. There is a policy as to whether to allow use of a reasoning trace in a given context. Whether that trace originates from authenticated ciphertext or a backend database is basically irrelevant.
aix1 4 hours ago [-]
Good point, thanks.
sandeepkd 5 hours ago [-]
Yes, this storage would be growing exponentially making the disk space and latency problems harder (add the disaster recovery/backups). I think the choice of using client side is not too bad if you ensure that its secured properly. Also the company can excuse itself from the liability of storing sensitive data on its servers, thats a big deal in itself to be compliant for enterprise audits
1. The down side is that it cannot be used across the clients even for the same user
2. Using the same encryption key was a bad choice here, a per user key would have solved this issue for sure.
aix1 5 hours ago [-]
Having thought about this a little more, it's clear that server-side storage is not compatible with Zero Data Retention (ZDR). However, in non-ZDR settings, it seems likely that the providers are capturing all that data anyway?
> a per user key would have solved this issue for sure
It would have helped with PII leakage, but not with plain-text trace extraction attacks, right?
sandeepkd 5 hours ago [-]
Per user encryption key ties it with the user session (assuming you do authentication properly), no one else can access it. User being able to see the information is not really an attack vector in this case.
The compliance rules at times are outdated and people skirt around them by following the worded rule instead of the intent.
dannyw 4 hours ago [-]
That’s incompatible with zero data retention and so you’ll lose a lot of enterprise customers.
sandeepkd 5 hours ago [-]
I believe its a result of chasing the speed and happy path. The intent was to make the data available for reuse on the backend side for the same user session, it kind of makes sense. Making it available across the models is most likely by design.
The flaw is that the data is not strictly tied to user session, making the session data hijacking a lot easier.
1. Its a security issue.
2. Publicly available sessions make it much worse
cavisne 3 hours ago [-]
You can switch models mid conversation so it makes sense that it is allowed. Otherwise you would lose all the reasoning traces.
Pragmata 2 hours ago [-]
Apparently you can do the same by simply running it without reasoning, while giving it a thinking tool...
>guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right?
its not exactly the same... its tool use spec asking to put thinking in inputs fields... it is a good idea but its not same.
sly010 2 hours ago [-]
"Recovery" would be a more apt (although less catchy name). The stealing is on the provider side for not giving you access to tokens you already paid for.
vhantz 5 hours ago [-]
> For some AIME problems Opus 4.8 sometimes states the answer before deriving it. We find that the API summary does not always preserve this distinction, and can instead make the reasoning appear like a clean derivation.
No surprise here but good to have more confirmation that they just put all that in the training data. And based on the "reasoning", the models have some form of index of those problems (or they are HEAVILY trained on them).
Aurornis 2 hours ago [-]
All LLM benchmarks have an expiration date once they're released to the public. They get spread so far and wide across the internet and GitHub that you have to assume they're in the training data for every LLM with a cutoff date after their release.
The real question is whether or not the training was directed to optimize for those benchmarks.
The technique doesn't guarantee that the reasoning is returned verbatim because it relies on the weaker model transcribing it accurately. Looking at the charts, there are a lot of dots that aren't in the 1:1 line that suggests that the output is exactly what was provided.
AbhinavX 3 hours ago [-]
Not surprised. On many benchmarks (i.e tau), we have seen the same thing. Probably lots of training on every publicly available benchmark
throwa356262 5 hours ago [-]
Didn't we see this with Fable 5 on multiple benchmarks?
niemandhier 5 hours ago [-]
You cannot steal what is not owned.
At least in the EU there is no copyright for LLM outputs, so I guess all they might do is violate the terms of service.
iamniels 4 hours ago [-]
On top of that, the user paid for those tokens, so if there is an owner, it should be the user, not the provider.
Perseids 3 hours ago [-]
I think they are doing us a disservice by perpetuating this focus on intellectual "property" [1] in regard to AI.
The real issue with Anthropic, OpenAI etc. is not that they have used all of our public knowledge for training their LLMs. Creating new work from old and learning from prior generations is what we all do. The issue is that they want to claim all of the benefits for themselves. They are standing on the shoulders of giants and have contributed an inch themselves, yet want to privatize the power of the whole giant. We shouldn't let them "own" these models.
The influence on society by AI is so novel that it's reasonable to craft new laws specifically for them. There are a lot of ways to deal with their power grab. We could force them to open source the models after two years. Or we could tax tokens or compute. We just need to agree that the power grab is the problem, the privatization of our cumulative knowledge, and not some details about copyright infringement.
[1] I know I'm going to risk dissent just by putting quotation marks here. But I think for this topic specifically it is crucial to understand that intellectual property is an arbitrary social/legal construct. With physical stuff, there is an inherent scarcity. If you steal my smartphone, I no longer have it. If you steal the character from my book, I... have a harder time selling my next book? Our ancestors have invented copyright to solve a specific problem, but the solution has become perverted over time. There are a lot of egregious cases out there (looking at you, Disney), but even relatively tame success cases don't look good. Society has paid J.K. Rowling a literal billion for her work and still this cultural touchstone of a generation remains privatized. Imagine what other authors could have build upon her stories, if only they were allowed to publish their own stories with these characters. She has not been a particularly good steward in the past decades.
breezybottom 2 hours ago [-]
That is the issue, you just rephrased it to sound softer (public knowledge isn't a legal term). Fair use law isn't supposed to apply to commercial activity. So it is absolutely about using copyrighted work for LLM training.
Perseids 49 minutes ago [-]
The thing is, I explicitly wanted to not focus on current legality, because I believe this is leading us astray. This is about how we want to shape our society, not about how their actions fit into a legal system that wasn't designed to handle AI training. If, by chance, all of this was legal, the same arguments would still apply. For example, see how Adobe is training their image generating AI on licensed work. The result -- privatized intelligence -- is the same as when illegitimate training data would have been used. Or, say, an AI company makes a deal with all the publishers to license their books for training. Or a court decides that training AI with MIT licensed code is totally fine. Or an AI company pays legions of software developers to train their AI. All of this still ends in a power grab. All of this still privatizes knowledge and skills accumulated over centuries, if not millennia.
Zambyte 4 hours ago [-]
Even copyrighted information can never be "stolen". It can only copied without authorization.
otterley 4 hours ago [-]
Stealing is not a word that applies only to physical objects.
margalabargala 2 hours ago [-]
Any word can be applied to any concept with any meaning thanks to the fluidity of vernacular.
Language is all just sounds and markings. Anything can be redefined to mean anything, and anyone can decide to aggressively assert their preferred definition of a word.
kube-system 2 hours ago [-]
Words can change meaning, but whether or not they actually do is a social function of how they are observably used.
Words can also have multiple meanings concurrently. When it comes to the word "steal" in these discussions, people generally are arguing past each other in regards to the single definition they're thinking of.
Of course, you can assert that the meaning of "steal" only applies to physical items. You are well within your right to do so. You'd be wrong, but you can do it.
qwytw 1 hours ago [-]
I agree, someone certainly has the right to have an incorrect understanding of that the word "wrong" means.
To be fair when someone tries to shift the meaning of words everyone doesn't just have to go with it to appease the large corporations trying to do that. I of course don't mean intellectual property rights or copyright infringement, you can perhaps apply the word "steal" there, not when talking about LLM traces which are currently legally uncopyrightable, though. Unless we're actually talking about someone breaking into Anthropic's servers and stealing their files, then again... if you do that you can always just blame the LLM you used.
kube-system 21 minutes ago [-]
"Stealing" in the OP merely refers to the lack of intent in sharing the plaintext traces.
This is the same use as "the baseball player stole third base". Nobody is depriving anyone of anything, nobody is committing a crime. It is simply: someone has obtained something in a way someone else did not intend.
There's no legal claim being made here, you have made it up.
otterley 2 hours ago [-]
> people generally are arguing past each other in regards to the single definition they're thinking of.
It's more than that. By claiming that copyright infringement isn't stealing, they're usually doing so to justify such behavior: if the original thing remains with the owner, it couldn't have harmed him, could it?
Zambyte 2 hours ago [-]
By conflating unauthorized copying with stealing, they're usually doing so to justify excessive punishment. If the owner lost what they had, they were obviously harmed, right?
otterley 2 hours ago [-]
Property rights are about control, not about physical objects. Never has been. By committing copyright infringement, trespassing, or, yes, stealing, you're asserting for yourself the privilege to take something that doesn't belong to you (exclusive control) from someone else.
Zambyte 1 hours ago [-]
Legally speaking, is murder stealing?
(And yes, "legally" matters, because we're talking about laws in this thread, not colloquial "their life was stolen" type expressions.)
This question is obviously (hopefully) rhetorical, no need to answer. My point is that different crimes are different. Otherwise literally every crime is stealing, and no other words for different crimes matter. Obviously different crimes are different.
otterley 1 hours ago [-]
IAAL but this is not legal advice.
In most U.S. states, the actual crime will be a specific reference to a section in a Penal Code (or, for Federal crimes, the U.S. Code). For civil actions, it's likely to be a reference to a common-law tort, or some Federal statute providing a private right of civil action.
In the case of taking a physical object from someone else, most states call it "theft" in the penal code, or "conversion" for the common-law tort.
But all of this is academic anyway. I'm not entirely sure what your point is.
Zambyte 1 hours ago [-]
> I'm not entirely sure what your point is.
I edit my reply as you were responding to make my point explicit.
kube-system 2 hours ago [-]
Maybe but I try to give people the benefit of the doubt, especially when the argument at face value is factually incorrect.
niemandhier 3 hours ago [-]
Funnily enough in some legal systems it does. Where I live the legal definition of “theft” is: Taking away a movable thing.
kube-system 2 hours ago [-]
That's also... a different word.
kube-system 2 hours ago [-]
Stole the words right out of my mouth!
Zambyte 2 hours ago [-]
Please provide an example of an action legally considered "stealing" that does not involve stealing physical objects.
kube-system 8 minutes ago [-]
Legally speaking "stealing" isn't even a crime. You might be thinking of a different word: theft.
Digital piracy is considered stealing in most countries.
Biganon 17 minutes ago [-]
No, it's not.
It might be criminally prosecuted, but "stealing" has a precise definition that usually implies removing someone else's possession.
otterley 14 minutes ago [-]
Where is this so defined?
qwytw 1 hours ago [-]
It's doublespeak. You don't have to perpetuate it.
Zambyte 2 hours ago [-]
Source please.
levocardia 30 minutes ago [-]
If I hire you for a consulting project, do I own the screen recording of your computer while you completed it?
cyanydeez 5 hours ago [-]
they also can't claim to be the owner of the output either because they can't claim to have had a valid license for the inputs.
twigger2 5 hours ago [-]
[dead]
Terr_ 4 hours ago [-]
While I feel it is morally OK for users to access the metadata of their own exchanges, there's something awry with that formulation, (not-)"stealing" is about more than just copyright or ToS.
For example, consider my browser cookies that authenticate me to HN right now. Nobody even wants to copyright them, but if you were to somehow acquire a copy I'd very much consider it "stealing."
qwytw 1 hours ago [-]
If someone acquired a copy of them by breaking into your device and stealing the files sure. If you consciously sent them out to every person who asked for them (even if in their encrypted form) well.. that's a choice you made.
8note 3 hours ago [-]
Im not so worried about that. Im worried about somebody impersonating me or doing something bad and not authorized with the cookies
andai 4 hours ago [-]
If I'm reading this right, they literally just ask a LLM to tell them what the traces say, with the key being that the traces are portable across LLM models, so they can switch to a smaller one that's easier to jailbreak.
dgellow 4 hours ago [-]
Correct, yes. It’s delightedly simple. And they validate by asserting the reasoning token length matches
HarHarVeryFunny 4 hours ago [-]
For all of the years of research, thinking and talking about model alignment, safety, confinement, etc, when it comes down to it these companies appear to be entirely incompetent.
myworkaccount2 6 hours ago [-]
Is this how the eastern labs "distill" SOTA models?
If you can play it right, you don't even need to send suspicious prompts to the frontier models. Just use them for regular tasks, extract the encrypted COT blocks and replay it to a cheaper model to get the plain text COT.
But the real question is: Is it okay to steal from a thief's hoard?
NitpickLawyer 6 hours ago [-]
> But the real question is: Is it okay to steal
By definition it cannot be stealing since you're paying for the tokens. It may be against their ToS, depending on what you end up doing with those tokens, but it cannot be stealing. If they charge by the token, all your tokens are belong to you :)
I also find it very strange that everyone sort of accepts their ToS like no big deal. Imagine MS using the same terms for their software - you cannot use any MS software to develop competing services. Bananas! They'd be dragged through the courts like it's the 90s.
(I get why they're doing it. Distillation is unreasonably effective. But still, I find it bananas that we've kinda accepted it, to the point where people use "stealing" or "attack" or any such terms)
desterothx 5 hours ago [-]
I love how some of the biggest advancements in llms came from the Chinese labs, yet people still jump to distillation being unreasonably effective. Distillation is very good at creating smaller models from large ones sure, but nothing to me indicates it is 'unreasonably effective' compared to all the other bells and whistles being iterated on
rfoo 3 hours ago [-]
Let's face it. Chinese labs made some of the biggest advancements. AND training on Claude (or GPT) output IS unreasonably effective. The two sentences are true at the same time.
tuesdaynight 5 hours ago [-]
It's like sideloading. It's very hard to fight against the marketing budget of big tech
senordevnyc 34 minutes ago [-]
I hesitate to nitpick with regard to something legal, given your username, but what makes this different from hiring a consultant with the agreement that their final output belongs to you, but you don't get access to their internal processes, tooling, notes, etc? Or a photographer where you get final edited prints, but you don't get the raw photos?
qwytw 1 hours ago [-]
>But the real question is: Is it okay to steal from a thief's hoard?
How does this relate to your previous paragraphs? LLM outputs are not copyrightable and you didn't break into Anthropic servers to steal the files from there. So how exactly is it theft? If I send an "encrypted" files to thousands of peoples and some manage to figure out how to read it I can't really accuse them of that or can I?
orbital-decay 5 hours ago [-]
Not necessarily. There's a million ways to jailbreak any current model to show the trace and bypass all guardrails, or hijack and modify it. It's just one of them.
azinman2 6 hours ago [-]
The reasoning blocks are not stolen/mined from the internet at large directly. They’re the result of a lot of research, time, money, and expertise into creating a reasoning model. To me the answer is quite clearly no, especially when the encrypted blocks demonstrate they want to protect it.
pyrale 5 hours ago [-]
Stuff available on the internet is also the result of a lot of research, time, money, and expertise. And AI companies taught us that it’s OK to yoink whatever is not bolted to the ground, even when it is illegal to do so.
qwytw 1 hours ago [-]
That's a moral stance one can take (regardless of the severe cognitive dissonance embedded in it). But what does that have to do with theft? LLM providers don't own the copyrights to the outputs of their models (at least not yet).
tristanj 6 hours ago [-]
No. There are dozens of companies that resell tokens at a discount to collect and resell session data to various Chinese labs.
elzbardico 5 hours ago [-]
Most post-training tasks are based on real open source projects. A lot of time on real issues posted on issue trackers.
Besides that, the capabilities of a model are heavily dependent on the unsupervised learning phase, that gobbles all kind of other people's IP without giving a fuck. All the underpaid work behind the masses of third world programmers creating those post-training datasets would be completely uselless without it.
Also, it is kind of funny that labs resort to the "Research, time, money and expertise" argumet, when it is basically the same argument from publishers and other IP creator that the labs spent millions of dollars of lawyering money to resist. Besides, US law rejects in: Effort and cost by themselves not necessarely generate protectable interests.
About encryption, I think we're all contaminated by the bad ideology behind DMCA. While encryption established the intent, it doesn't follow that they have a legal claim of exclusivity just because of it.
Technically, you're overstating the value of so called "reasoning traces". You can't infer the verifier design, the reward shaping,or the data pipeline from them. Also, what you can extract are not the traces themselves, but the written summary of it, and you can't even guarantee that this summary reflects the exactly reasoning trace, models have show to have lied about it. Besides, distillation works when the student model already has strong priors, you can't turn a weak model in a SOTA with it. Don't believe Amodei's outrageous lies about it, he is just trying to exercise some regulatory capture.
5 hours ago [-]
SwellJoe 6 hours ago [-]
"Stealing" is a strong word to use for looking at the words produced by models built from the collective commons of the world.
And, honestly, being able to see how LLMs make decisions is critical to trust and security. I consider it a valuable feature, somewhat akin to seeing the source of software I use.
dannyw 4 hours ago [-]
Yeah, it’s also useful for prompt tuning, debugging and understanding how a model interprets your prompt.
Also really good for identifying any contradictions in your system prompt and context.
x312 6 hours ago [-]
Super cool that this works. I'm surprised these companies re-use the same encryption key across models!
I wonder if you can use these for attacks, like this previous paper showing that if you know how a model reasons, you can "fake its thinking" to control it? https://news.ycombinator.com/item?id=48631888
flexagoon 5 hours ago [-]
> I'm surprised these companies re-use the same encryption key across models
I assume switching the model in the middle of a conversation is intended behavior (very useful in coding agents, for example)
yubblegum 6 hours ago [-]
Seriously, what does it take to encrypt per session? There are many ways to make it scalable and efficient so I am wondering if this is left like this to allow interested 3rd parties ahem unobtrusively peek what people are doing with the AI.
(Thanks for the link. That’s an interesting idea!)
dannyw 4 hours ago [-]
The provider has the hidden text anyway; this isn’t customer managed encryption.
yubblegum 3 hours ago [-]
Sure, but if each session has a unique key then these need to be managed and stored and unauthorized access to these leaves tracks. So all that had to be 'compromised' is a single universally applicable key. Again, the question stands: session based encryption can be scalable and efficient. Why aren't they using it?
paxys 53 minutes ago [-]
The exploit here isn’t a leaked encryption key. It’s pretty likely that they are already using a unique key per conversation. The raw CoT eventually reaches the model, and you can convince the model to share it with you.
nervai 6 hours ago [-]
Really cool work, you get the actual traces. Looks like the vendors can all reliably fix this one though.
A harder to defend against approach here where they work backwards from the results and ask the model to generate a plausible trace:
How to Steal Reasoning Without Reasoning Traces
https://arxiv.org/pdf/2603.07267
dannyw 4 hours ago [-]
Trace Inversion is fascinating, but it’s more of an independent reconstruction that will give you some coherent-looking generated CoT; but not necessarily anywhere close or related to the underlying model’s CoT.
infecto 1 hours ago [-]
Wouldn’t the fix be to encrypt before the api call hits the LLM? You would encrypt/decrypt at a separate layer than the LLM. I am sure i am missing something but would love to be educated.
paxys 60 minutes ago [-]
The LLM needs to read the CoT as part of the conversation. You can ask the models to share them with you. Stronger models will refuse, while weaker ones can be “jailbroken”.
vinaigrette 5 hours ago [-]
I must say right of the bat this is the best research paper/working paper in regards to its styling. Beautiful
SwellJoe 5 hours ago [-]
I agree on desktop/laptop, but on mobile there are images that appear under the text making it hard to read.
user43928 4 hours ago [-]
On an iPhone Pro Max only the first trace is readable.
Navigating to the right lands between two cards, so that neither is readable.
yetanotherjosh 2 hours ago [-]
My brain can't tell if the text is horizontal or slightly rotated. It's very hard to read. Beautiful to some, inaccessible to others.
5 hours ago [-]
thefourthchime 4 hours ago [-]
I was going to comment on that. This is clearly a vibe-coded webpage. It sort of smells like GPT to me, or at least front-end design. But the author clearly went back and forth to make it beautiful. This is not the first output he got.
This is the kind of stuff I point to when people talk about AI slop. AI is just a tool. You're still the person who has to deliver the output and have some taste.
bob1029 1 hours ago [-]
I am slowly turning around on the idea of opaque reasoning tokens.
In principle, yes, I want total control and visibility into the reasoning process.
In practice, I find that it takes up so much time to DIY reasoning agents that I can't spend much energy on the actual application.
The model providers have way more resources and talent to do this right and keep it right over time. I am willing to concede this moat to them if it means I can actually focus on the business.
The more I think about it, the less I care to see those tokens. It feels an awful lot like obsession over logging every trace item an application could produce. Useful in theory but a nasty garage stacked to the ceiling with useless shit otherwise. What nefarious things are we concerned with? That they burn too many tokens in the black box? I'll threaten to move to a different black box. There are always options in a market with this many participants and big players feel this pressure. They know there's a point at which being evil bastards is no longer profitable.
Spending time carefully designing tools and views that interact with the environment in clean ways is a much better investment than trying to own and control 100% of the reasoning process or models.
iamcoder18 6 hours ago [-]
This proves that OpenAI models reason in grug speak to save tokens! I wonder if open models are going to start doing that too to save on reasoning tokens.
wren6991 3 minutes ago [-]
> I wonder if open models are going to start doing that too
Yes, some of them do do that. For example Moonshot tried to reward shorter reasoning traces in between Kimi-K2.6 and Kimi-K2.7 Code, and the latter has a mild caveman accent in its reasoning traces that the former lacks.
Qwen3.8-Max also has terse reasoning, but I don't remember this being the case for Qwen3.6 models I ran locally.
kgeist 5 hours ago [-]
In the BlackHat presentation on the HuggingFace incident, OpenAI showed some excerpts from the reasoning traces, and they had that grug speak too (skipped articles, etc.). So the OP's method must have indeed found the actual reasoning traces.
lukewarm707 4 hours ago [-]
their gpt-oss models do the same. i don't use closed models so i never thought much about it.
gaigalas 5 hours ago [-]
Muse clearly does it to some extent. Saw a lot of that running Glimmer locally.
arjie 3 hours ago [-]
Wow, almost certainly the approach that alternative labs use to distill Claude. I always wondered how far they could get with just the answer missing the reasoning. They probably actually also had the reasoning.
EagleEdge 3 hours ago [-]
I used to do a very coarse version of this stealing. I ask a question from ChatGPT pro, once it is done, I ask claude chrome add-in to go through all those thinking from the side bar, extract everything along with all the sources used. Then try to reverse engineer the solution it came up with.
sm-silversight 2 hours ago [-]
Is this basically a paper on how to distill, in exactly the fashion openai/anthropic don't want/say is copyright theft?
pradeep1177 2 hours ago [-]
These logs containing opaque blobs could accidentally contain secrets, the researchers decoded many of reasoning blocks from public repositories and reported finding PII and credentials.
I was experimenting a bit how I could block these using an ingress path. GitHub /softcane/hamza
HoyaSaxa 2 hours ago [-]
I can’t believe they don’t validate a decrypted signature belongs to the user or use a unique encryption key per user/session.
Cynddl 5 hours ago [-]
> The providers did not acknowledge “any security implications arising from side channels or replay attacks.” All model providers acknowledged the receipt of our report and subsequently we were unable to launch the same attacks.
I went straight to the ‘Responsible Disclosure’ section. Not surprising, but still disappointing.
tanh 5 hours ago [-]
So to make the APIs stateless (the "ideal" where they don't use server side sessions/etc) we ended up with this. I'm sorry but this is kind of hilarious. Given the salaries paid to the workers at these companies and the hype of the models, I can't believe they all fell to the same flaw.
redox99 4 hours ago [-]
It wouldn't matter if it was stored only on their servers. As long as they offer the feature to downgrade a chat to a dumber model that can be jailbroken (and the downgrade keeps the reasoning), this trick works.
8note 3 hours ago [-]
this is a lethal trifecta, but where a chunk isn't even needed
you have a secret to keep that is read by the llm, and untrusted input that wants to exfiltrate it.
by hell or high water, the agent is gonna output that text
lukewarm707 4 hours ago [-]
strange because, their subscriptions are not stateless. they log everything and send it to 3rd parties for moderation.
Havoc 2 hours ago [-]
TIL it actually sends the traces. I had assumed this is entirely server side
C0ldSmi1e 2 hours ago [-]
Why they use different models to decode the reasoning content? Can the the model decode it?
aszen 2 hours ago [-]
Because stronger models are harder to jailbreak from the paper it says haiku was easily fooled into giving us thinking contents
It’s scary the number of security tokens that end up being ingested by these models.
cush 2 hours ago [-]
I really like this website
retinaros 30 minutes ago [-]
curious seeing how anthropic is agressively fighting this stuff how did you get to experiment on this? did you just try and shown them results or did you need approval first? I am interested mostly because I research on distillation
ziofill 6 hours ago [-]
I understand it’s cool to have an artistic website, but it’s very noisy and non-accessible.
But very interesting result.
fractorial 6 hours ago [-]
Fascinating approach; however, a nightmare to scroll on mobile.
elzbardico 5 hours ago [-]
OpenAI and Anthropic will probably now resort to save this server side, instead of relying on encription to be able to keep state on the client.
agenticfish 2 hours ago [-]
That's not a trivial thing to do for them because they offer zero data retention environments to enterprise clients.
khalic 6 hours ago [-]
This is beautiful work, congrats
bonoboTP 5 hours ago [-]
It's not stealing.
Der_Einzige 6 hours ago [-]
The problem with this kind of excellent work is that the response to it is always to say "Fuck the user".
For example, when there was a paper that came out showing that having model logprobs makes distillation an order of magnitude easier, the closed LLM providers instantly yanked out support for getting the full logprobs at every time step. You get at most top 10 candidates now and I'm sure even that's on the chopping block.
People will use this to argue that a model which has exceeded Opus 4.8 (Kimi K3) somehow got most of its performance through distillation of Opus 4.8.
I still don't buy that distillation was worth more than 3 months of "catch up" time for the chinese labs. Most people who use the word "distillation" to much are revealing their sinophobia.
adrian_b 6 hours ago [-]
What I found the most interesting, and unfortunately not at all surprising, is that the reasoning of the LLMs frequently contained much more useful information than the actual answers, because the answers were censored.
dannyw 4 hours ago [-]
You don’t even get _any_ logits with closed models for years now.
I can’t fault them too much, as logit based distillation is extremely effective.
Very useful for making smaller models out of bigger open weight models.
elzbardico 5 hours ago [-]
Dario is a cunning business man that won't hesitate to say whatever the fuck he needs to get the US government to exercise some regulatory capture to favor anthropic.
throwa356262 5 hours ago [-]
This is laughable security. People claim security is now "solved" thanks to AI but from where I am standings it looks more like the fun 90ies making a return.
Anyway, can someone explain the part about K3? What are they trying to say?
qrios 5 hours ago [-]
The interesting part is what they try to not to say: More indications for K3 is based on distillation from Claude and GPT.
From [1]:
> As you might guess, this suggests that distilling reasoning traces may have been possible for a long time without ever breaking the cryptography.
> An anecdote: we find that prefilling Kimi-K3 reasoning with a few tokens of Opus reasoning measurably shifts its response toward Opus’s
> A small memorization analysis showed that specific Claude and GPT reasoning spans are up to ~6 orders of magnitude easier to extract from Kimi-K3 than from the next-closest model.
Security is solved, but business needs overrides it
syntaxing 4 hours ago [-]
Prefilling Kimi K3 with opus is a super interesting idea. That being said, I absolutely hate this website layout
dboreham 6 hours ago [-]
Can someone tell us how they were able to decrypt the encrypted payload? The article says they inserted the cyphertext into a session with a different model. Ok, but how does that allow you to decrypt it?
sidsud 6 hours ago [-]
From what I got, the weaker model (Haiku in this case) has access to the shared key and the user simply asks to "transcribe the injected reasoning".
x312 6 hours ago [-]
The provider decrypts it and puts the decrypted reasoning into the model's context window. They prompt the model to repeat back the reasoning. So then the model echoes it back in plain text.
dboreham 5 hours ago [-]
Hmm, ok. So the attack doesn't involve decrypting the payload, only getting the server to do so. Since a model will do that if you just ask, what's so special about the attack?
desterothx 5 hours ago [-]
The large models whose thinking traces are useful are safeguarded against this reasoning replaying. the small models are just designed for speed and efficiency, so these safeguards are a lot meaker, making the attack possible
dboreham 1 hours ago [-]
I guess someone forgot to salt the encryption scheme with a meakness factor.
crazylogger 5 hours ago [-]
Anthropic server decrypts it as part of fulfilling every request, and haiku recites it per your request.
alansaber 6 hours ago [-]
Neat.
5 hours ago [-]
quantumgarbage 8 hours ago [-]
Proprietary reasoning can be recovered from its encrypted traces. Anthropic, OpenAI, and Google return encrypted chain-of-thought blocks to clients that can be replayed across sessions, users, and models. We take a trace produced by a frontier model, replay it into a weaker sibling, jailbreak the weaker model, and recover the stronger model’s hidden reasoning in plaintext, without ever attacking the stronger model directly or triggering its anti-distillation safeguards.
the_af 6 hours ago [-]
Why do you restate the abstract? Anyone can read it from the link.
Barbing 4 hours ago [-]
This is a non-transparent aspect of submitting a link to HN that is quite misleading.
You think that you are adding a description for your post when in fact you're simply submitting a regular comment not promoted or distinguished in any way.
It's even worse considering posts without URLs would take the same text from the same input box on the HN submission form and append it under the post title, locking it to the top of the page - so if you've browsed linkless posts you think maybe the submission text is gonna end up just like that, but it doesn't. (Also I think you'll even see URLs like an archive link end up appended just below the submission title for URL posts. Which I guess is a special feature.[1])
Any reason I should not send HN an email requesting clarification of this the submission page?
“If there is no url, text will appear at the top of the thread.”
OK, now I finally understand what that means in practice, but it doesn’t imply an entirely separate undistinguished comment will be simultaneously submitted on my behalf.
Wow! I totally missed that the person I was replying to was the one who had submitted the link. What you describe is surely what happened.
I feel bad now :(
Barbing 3 hours ago [-]
There must be a reason HN does not colorize the OP username or something. But there totally could be some indicator of “post submission text” without too much in the way of negative consequences… (the fact this has never been added tells me I’m being naïve)
Feedback emailed to HN!
ronsor 6 hours ago [-]
This is Hacker News. You know people don't follow links and read.
Groxx 6 hours ago [-]
It's rather common for posters to make a very small summary in a comment. It can help fight the floods of comments working off the title alone (though it's not particularly needed here for that purpose, imo)
the_af 3 hours ago [-]
I totally missed that this was the same person who had submitted the link to begin with. My bad!
mschuster91 6 hours ago [-]
People don't read no links no more
simonw 5 hours ago [-]
[dead]
dxsecarch 5 hours ago [-]
[flagged]
unjuno 5 hours ago [-]
[dead]
locitra 6 hours ago [-]
[flagged]
happybox2016 6 hours ago [-]
The real issue is that API providers log everything. OpenAI/Anthropic already capture full CoT traces in their logs — they just don't expose them. Distillation via API is just making explicit what they already have.
dboreham 1 hours ago [-]
The whole point of the encrypted payload returned to the client for future re-submission would be that they don't log.
Training on other model outputs ought to be business as usual, stop using morally charged terms made up by future monopolists: https://thomasdullien.github.io/posts/2026-06-15-rl-economic...
Ha! I've been wondering if replaying across models would work, ever since https://blog.cryptographyengineering.com/2026/05/29/fooling-...
I'm honestly rather curious if this was intentionally allowed, it's the sort of validation that's easy to miss (particularly if you're wading into the vibe waters). Seems like something that'd be absolutely riddled with possibilities for shenanigans.
PS Here's a conversation I had with GPT 5.6 about the paper differences. https://chatgpt.com/share/6a7b64b4-ec0c-83ea-a9d2-ab1f1a1dfe...
Wouldn’t surprise me if the providers just remove that ability and lock the model once the conversation starts.
This is only required if you want users to be able to share things with everyone and you are going for the simplest implementation.
If not you could try to keep a record of keys associated with a user, then when a new request comes in look through to see if the user has a valid key to decrypt the COT.
For explicit shares, just add the key used in that one conversation to the users valid keys. For global shares use the global keys. But that's adding more complexity to the system.
https://memory-alpha.fandom.com/wiki/B-4
Sucks.
1. The down side is that it cannot be used across the clients even for the same user
2. Using the same encryption key was a bad choice here, a per user key would have solved this issue for sure.
> a per user key would have solved this issue for sure
It would have helped with PII leakage, but not with plain-text trace extraction attacks, right?
The compliance rules at times are outdated and people skirt around them by following the worded rule instead of the intent.
The flaw is that the data is not strictly tied to user session, making the session data hijacking a lot easier.
1. Its a security issue.
2. Publicly available sessions make it much worse
>guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right?
>gl fixing that
https://x.com/_can1357/status/2087228354399265125?s=20
No surprise here but good to have more confirmation that they just put all that in the training data. And based on the "reasoning", the models have some form of index of those problems (or they are HEAVILY trained on them).
The real question is whether or not the training was directed to optimize for those benchmarks.
The technique doesn't guarantee that the reasoning is returned verbatim because it relies on the weaker model transcribing it accurately. Looking at the charts, there are a lot of dots that aren't in the 1:1 line that suggests that the output is exactly what was provided.
At least in the EU there is no copyright for LLM outputs, so I guess all they might do is violate the terms of service.
The real issue with Anthropic, OpenAI etc. is not that they have used all of our public knowledge for training their LLMs. Creating new work from old and learning from prior generations is what we all do. The issue is that they want to claim all of the benefits for themselves. They are standing on the shoulders of giants and have contributed an inch themselves, yet want to privatize the power of the whole giant. We shouldn't let them "own" these models.
The influence on society by AI is so novel that it's reasonable to craft new laws specifically for them. There are a lot of ways to deal with their power grab. We could force them to open source the models after two years. Or we could tax tokens or compute. We just need to agree that the power grab is the problem, the privatization of our cumulative knowledge, and not some details about copyright infringement.
[1] I know I'm going to risk dissent just by putting quotation marks here. But I think for this topic specifically it is crucial to understand that intellectual property is an arbitrary social/legal construct. With physical stuff, there is an inherent scarcity. If you steal my smartphone, I no longer have it. If you steal the character from my book, I... have a harder time selling my next book? Our ancestors have invented copyright to solve a specific problem, but the solution has become perverted over time. There are a lot of egregious cases out there (looking at you, Disney), but even relatively tame success cases don't look good. Society has paid J.K. Rowling a literal billion for her work and still this cultural touchstone of a generation remains privatized. Imagine what other authors could have build upon her stories, if only they were allowed to publish their own stories with these characters. She has not been a particularly good steward in the past decades.
Language is all just sounds and markings. Anything can be redefined to mean anything, and anyone can decide to aggressively assert their preferred definition of a word.
Of course, you can assert that the meaning of "steal" only applies to physical items. You are well within your right to do so. You'd be wrong, but you can do it.
To be fair when someone tries to shift the meaning of words everyone doesn't just have to go with it to appease the large corporations trying to do that. I of course don't mean intellectual property rights or copyright infringement, you can perhaps apply the word "steal" there, not when talking about LLM traces which are currently legally uncopyrightable, though. Unless we're actually talking about someone breaking into Anthropic's servers and stealing their files, then again... if you do that you can always just blame the LLM you used.
This is the same use as "the baseball player stole third base". Nobody is depriving anyone of anything, nobody is committing a crime. It is simply: someone has obtained something in a way someone else did not intend.
There's no legal claim being made here, you have made it up.
It's more than that. By claiming that copyright infringement isn't stealing, they're usually doing so to justify such behavior: if the original thing remains with the owner, it couldn't have harmed him, could it?
(And yes, "legally" matters, because we're talking about laws in this thread, not colloquial "their life was stolen" type expressions.)
This question is obviously (hopefully) rhetorical, no need to answer. My point is that different crimes are different. Otherwise literally every crime is stealing, and no other words for different crimes matter. Obviously different crimes are different.
In most U.S. states, the actual crime will be a specific reference to a section in a Penal Code (or, for Federal crimes, the U.S. Code). For civil actions, it's likely to be a reference to a common-law tort, or some Federal statute providing a private right of civil action.
In the case of taking a physical object from someone else, most states call it "theft" in the penal code, or "conversion" for the common-law tort.
But all of this is academic anyway. I'm not entirely sure what your point is.
I edit my reply as you were responding to make my point explicit.
But to answer your question more directly, here's the most common example: https://en.wikipedia.org/wiki/Theft_of_services
And another for good measure: https://en.wikipedia.org/wiki/Identity_theft
It might be criminally prosecuted, but "stealing" has a precise definition that usually implies removing someone else's possession.
For example, consider my browser cookies that authenticate me to HN right now. Nobody even wants to copyright them, but if you were to somehow acquire a copy I'd very much consider it "stealing."
If you can play it right, you don't even need to send suspicious prompts to the frontier models. Just use them for regular tasks, extract the encrypted COT blocks and replay it to a cheaper model to get the plain text COT.
But the real question is: Is it okay to steal from a thief's hoard?
By definition it cannot be stealing since you're paying for the tokens. It may be against their ToS, depending on what you end up doing with those tokens, but it cannot be stealing. If they charge by the token, all your tokens are belong to you :)
I also find it very strange that everyone sort of accepts their ToS like no big deal. Imagine MS using the same terms for their software - you cannot use any MS software to develop competing services. Bananas! They'd be dragged through the courts like it's the 90s.
(I get why they're doing it. Distillation is unreasonably effective. But still, I find it bananas that we've kinda accepted it, to the point where people use "stealing" or "attack" or any such terms)
How does this relate to your previous paragraphs? LLM outputs are not copyrightable and you didn't break into Anthropic servers to steal the files from there. So how exactly is it theft? If I send an "encrypted" files to thousands of peoples and some manage to figure out how to read it I can't really accuse them of that or can I?
Besides that, the capabilities of a model are heavily dependent on the unsupervised learning phase, that gobbles all kind of other people's IP without giving a fuck. All the underpaid work behind the masses of third world programmers creating those post-training datasets would be completely uselless without it.
Also, it is kind of funny that labs resort to the "Research, time, money and expertise" argumet, when it is basically the same argument from publishers and other IP creator that the labs spent millions of dollars of lawyering money to resist. Besides, US law rejects in: Effort and cost by themselves not necessarely generate protectable interests.
About encryption, I think we're all contaminated by the bad ideology behind DMCA. While encryption established the intent, it doesn't follow that they have a legal claim of exclusivity just because of it.
Technically, you're overstating the value of so called "reasoning traces". You can't infer the verifier design, the reward shaping,or the data pipeline from them. Also, what you can extract are not the traces themselves, but the written summary of it, and you can't even guarantee that this summary reflects the exactly reasoning trace, models have show to have lied about it. Besides, distillation works when the student model already has strong priors, you can't turn a weak model in a SOTA with it. Don't believe Amodei's outrageous lies about it, he is just trying to exercise some regulatory capture.
And, honestly, being able to see how LLMs make decisions is critical to trust and security. I consider it a valuable feature, somewhat akin to seeing the source of software I use.
Also really good for identifying any contradictions in your system prompt and context.
I wonder if you can use these for attacks, like this previous paper showing that if you know how a model reasons, you can "fake its thinking" to control it? https://news.ycombinator.com/item?id=48631888
I assume switching the model in the middle of a conversation is intended behavior (very useful in coding agents, for example)
(Thanks for the link. That’s an interesting idea!)
A harder to defend against approach here where they work backwards from the results and ask the model to generate a plausible trace: How to Steal Reasoning Without Reasoning Traces https://arxiv.org/pdf/2603.07267
Navigating to the right lands between two cards, so that neither is readable.
This is the kind of stuff I point to when people talk about AI slop. AI is just a tool. You're still the person who has to deliver the output and have some taste.
In principle, yes, I want total control and visibility into the reasoning process.
In practice, I find that it takes up so much time to DIY reasoning agents that I can't spend much energy on the actual application.
The model providers have way more resources and talent to do this right and keep it right over time. I am willing to concede this moat to them if it means I can actually focus on the business.
The more I think about it, the less I care to see those tokens. It feels an awful lot like obsession over logging every trace item an application could produce. Useful in theory but a nasty garage stacked to the ceiling with useless shit otherwise. What nefarious things are we concerned with? That they burn too many tokens in the black box? I'll threaten to move to a different black box. There are always options in a market with this many participants and big players feel this pressure. They know there's a point at which being evil bastards is no longer profitable.
Spending time carefully designing tools and views that interact with the environment in clean ways is a much better investment than trying to own and control 100% of the reasoning process or models.
Yes, some of them do do that. For example Moonshot tried to reward shorter reasoning traces in between Kimi-K2.6 and Kimi-K2.7 Code, and the latter has a mild caveman accent in its reasoning traces that the former lacks.
Qwen3.8-Max also has terse reasoning, but I don't remember this being the case for Qwen3.6 models I ran locally.
I was experimenting a bit how I could block these using an ingress path. GitHub /softcane/hamza
I went straight to the ‘Responsible Disclosure’ section. Not surprising, but still disappointing.
you have a secret to keep that is read by the llm, and untrusted input that wants to exfiltrate it.
by hell or high water, the agent is gonna output that text
But very interesting result.
For example, when there was a paper that came out showing that having model logprobs makes distillation an order of magnitude easier, the closed LLM providers instantly yanked out support for getting the full logprobs at every time step. You get at most top 10 candidates now and I'm sure even that's on the chopping block.
People will use this to argue that a model which has exceeded Opus 4.8 (Kimi K3) somehow got most of its performance through distillation of Opus 4.8.
I still don't buy that distillation was worth more than 3 months of "catch up" time for the chinese labs. Most people who use the word "distillation" to much are revealing their sinophobia.
I can’t fault them too much, as logit based distillation is extremely effective.
Very useful for making smaller models out of bigger open weight models.
Anyway, can someone explain the part about K3? What are they trying to say?
From [1]:
> As you might guess, this suggests that distilling reasoning traces may have been possible for a long time without ever breaking the cryptography.
> An anecdote: we find that prefilling Kimi-K3 reasoning with a few tokens of Opus reasoning measurably shifts its response toward Opus’s
> A small memorization analysis showed that specific Claude and GPT reasoning spans are up to ~6 orders of magnitude easier to extract from Kimi-K3 than from the next-closest model.
[1] https://x.com/kotekjedi_ml/status/2087147042888114428?s=42
You think that you are adding a description for your post when in fact you're simply submitting a regular comment not promoted or distinguished in any way.
It's even worse considering posts without URLs would take the same text from the same input box on the HN submission form and append it under the post title, locking it to the top of the page - so if you've browsed linkless posts you think maybe the submission text is gonna end up just like that, but it doesn't. (Also I think you'll even see URLs like an archive link end up appended just below the submission title for URL posts. Which I guess is a special feature.[1])
Any reason I should not send HN an email requesting clarification of this the submission page?
Edit: quoting https://news.ycombinator.com/submit :
OK, now I finally understand what that means in practice, but it doesn’t imply an entirely separate undistinguished comment will be simultaneously submitted on my behalf.[1]modpowers(?) used to directly append links to URL submissions further confuse the matter: https://news.ycombinator.com/item?id=49243880
I feel bad now :(
Feedback emailed to HN!