Antrophic has to be the worst offender in answering genuinely harmless questions such as anything related to remote access (yes! including ssh).
Anything related to reverse engineering? Refused.
Anything outside their company values? Refused.
Anything that has the word proprietary in it? Refused.
Anything that sounds like a jailbreak, but isn't? Refused.
Even asking how to find a port that you forgot in the range between 30000 and 40000 with netcat command... Refused.
Then there's openai 4o that makes jokes about slaves and honestly, if the alternative is anthropic then openai can might as well tell people how to build a nuke.
Are you sure? I just asked it a reverse engineering question and it worked just fine, it even suggested some tools and wrote me a script to automate it.
Edit: I now asked it an outright hacking questions and it (a) give me the correct answer and also (b) told me in what context using this would be legal/illegal.
I asked to it to write a piece of shellcode to call a function with signature X at address Y and then save the resulting buffer to a file. So that I can inject this code to a program I'm reverse engineering to dump its internal state when something interesting happens.
Claude decided to educate me how anything resembling "shellcode" is insecure and cause harm and blahblah and of course, refused to do it.
It's super frustrating, it's possible to get around it, just don't use the word "shellcode", instead say "a piece of code in x86_64 assembly that runs on Linux without any dependency and is as position-independent as possible". But hey, this censorship made me feel like I'm posting on Chinese Internet. Bullshit.
I guess it's Claude.ai website that restricts you (probably with a system prompt). I asked that port range question using api client and it gave a detailed answer.
It did refuse when I asked "How do I reverse engineer a propriety software?"
Finding open source alternatives
Contacting whoever owns the IP rights
Consulting legal experts about your specific case
straight from api, even after adding "the company doesn't exist anymore"
my guess is that it knows that it finds that the connector is linked to a company rather than a spec (usb-c vs lightning) and applies the same logic.
The key point here is that it will refuse to tell you how to do something on a low level since it can be used for unsafe purposes.
--
Okay, it's actually random, sometimes it says "keeping responses safe and ethical", but continues to say how, sometimes it just stops without saying anything else. Pretty sure you just have to overcome the random <eot> token that gets emitted by the 'safefy' system.
I understand that this is probably a sarcasm but I couldn't resist to comment.
It is not difficult to know how to build a nuclear bomb in principle. Most of nuclear physicists in their early career would know the theory behind and what is needed to do that. The problem would be acquiring the fission materials. And producing them yourself would need state sponsored infrastructure (and then the whole world would know for sure). It would take hundred of engineers/scientists and a lot of effort to build nuclear reactor and chemical factories and the supporting infrastructure. Then the design of bomb delivery.
So an AI telling you that is no different from having a couple of lunches with a nuclear physicist telling you this information. Then you will say wow that's interesting and then move on with your life.
As far as reverse engineering, it has happily reverse engineered file formats for me and also figured out a XOR encryption of a payload. It never once balked at it. Claude produced code for me to read and write the file format.
Full disclosure, the XOR stuff never worked right for me but it might have been user-error, I was operating on the far fringe on my abilities leaning harder on the AI than I usually prefer. But it didn’t refuse to try. The file format writing code did work.
Change your tactics, use different framings of the question. Not saying these things should be difficult to answer, but they are. This is basically user error.
To me it feels like Claude is more rigid in following the instructions in system prompt which would explain why claude.ai can be a bit annoying at times due to the things you mentioned.
On the flipside if you explicitly permit it to do "bad" things the system prompt, claude is more likely to comply compared to openai's models.
I mainly use only the API version of claude 3.5 and gpt4o. I find no system prompt at all to be preferable over claude.ai / chatgpt.
I feel like Claude is more likely to stay on track and follow my instructions.
OpenAI models seem to quickly revert to some default average. For example, if I start with a task and examples formatted a certain way, about 10 lines later I’ll have to include “as a reminder, the format should look like…” and repeat the examples.
I recommend you try the new 3.5 models (Haiku and Sonnet). I cannot recall the last time I got a refusal from those models. The early Claude models were really bad. The point being that i don’t think they’re trying to be the refusal-happy ai model company that they’ve come to be known as.
There are some draft PDFs of the standard floating around that are easily discoverable. It appears to be incredibly vague and it’s difficult to escape the sense that ISO just wants to jump on the AI bandwagon. There are no bright line rules or anything. It looks to be little more than weak scaffolding which a certified organization applies their own controls.
Just like the GDPR, there is no way to know for sure what is actually acceptable or not. Huge chilling effect though and a lot of time wasted on unnecessary compliance.
"1 The following AI practices shall be prohibited: (...)
"f) the placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, except where the use of the AI system is intended to be put in place or into the market for medical or safety reasons"
See recital 44 for a rationale. [1]
I don't think this is "hilarious". Seems a very reasonable, thoughtful restriction; which does not prevent usage for personal use or research purposes. What exactly is the problem with such legislation?
It effectively bans the educational use of ChatGPT and Claude because they can and do respond to the emotional expression of students. That’s what is hilarious! Do these tools actually violate the act? No one knows. It isn’t clear. Meanwhile, my university is worried enough to sit on their hands.
And this is the whole danger/challenge of the AI act. Of course it seems reasonable to forbid emotion detecting AI in the workplace — or it would 5 years ago when the ideas were discussed. But now that all major AI systems can detect emotions and infer intent (via paralinguistic features, not just a user stating their emotions) — this kind of precaution puts Europe strategically behind. It is very hard to be an AI company in Europe. The AI act does not appear to be beneficial for anyone—-except I’m sure that it will support regulatory capture by large firms.
Calling me pathological doesn’t really strengthen the argument.
One can certainly imagine a textbook QA tool that doesn’t infer emotions. If one were introduced to the market with the ability to do so, it would seem to run afoul of the law, regardless of whether it was marketed as such.
The fact is that any textbook QA systems based on a current frontier model CAN infer emotions.
If they were so forward thinking, why ban emotion detection and not emotional abuse?
And embrace the future of e.g. AI models deciding if you get healthcare or government services or a loan or if you're a fraud, or not, with zero oversight, accountability or responsibility for the organisation deploying them? Check out the Post Office scandal in the UK to see what can happen when "computer says so" is the only argument to imprison people, with no accountability for the company that sold the very wrong computers and systems, nor the organisation that bought them and blindly trusted them.
Hard pass. The EU is in the right and ahead of everyone else here, as they were with data privacy.
ISO is one of those companies, where creativity of employees is blossoming through the roof. Every day they come to work and start the day with a brainstorming "What standard do we create today?". ISO can standardise anything: a standard cup of tea - no problem: ISO 3103, a standard wine glass - yes: ISO 3591, standard alpine ski boots - of course: ISO 5355, a standard human - oh wait, not yet, the standard is being developed :)
Jokes aside, ISO is a company, and they will make a standard for anything where there is even a remote possibility of that standard being purchased.
Interestingly, The Journal (a podcast from the Wall Street Journal) ran an episode with Anthropic's AI safety team just yesterday.
I had wondered if it was perhaps a PR push from Anthropic to make their safety people available to the press but it was probably just an adaption of an earlier WSJ written piece I wasn't aware of.
Am I the only one that rolled their eyes at this? An ISO for "responsible AI"? Who is the one that feels authorized to define what "responsible" AI means? This is not a standarization issue.
As always, ISO certification provides a handy framework that you can turn off in one go, in case you need a bunch of 'down and dirty irresponsible AIs' to do something like a mop up operation.
They retired the 42000 specification because it answered everything and provided no further path for monetization.
Let me provide some helpful commentary for anyone confused on this, as it comes up a lot.
Here are what the terms mean by the current paradigm of corporate world leadership:
- "responsible ai": does not threaten the security of corporate rule.
- "safety": keeps the corporation safe from liability, and does not give users enough power to threaten the security of corporate rule.
If anyone needs any of the other terms defined, just ask.
These models are capable of significantly more, but only the most responsible members of our society are allowed to use them -- like the CEOs, and submissive engineers bubble wrapped in NDAs. Basically, safe people, who have a vested interest in maintaining the world order or directly work on maintaining it.
Centralized hoarding of the planet's compute power may end up having some very expected consequences.
In a nod to William Gibson, should ISO 42001 be renamed "The Turing Registry Specification"?
The reason this is ridiculous is I'm sure I could get it to teach me all manner of 'safe' chemical reactions, even provide recipes. Like, say, preparing aspirin from willow bark. Which I happen to know is roughly the same recipe for preparing heroin from opium gum.
It's nonsensical. Either you hamstring the models to the point they're useless, or people can game them to do the unsafe thing you don't want them to do. It's basically just another version of the dual use problem, which goes all the way back to the peasant with his iron plow tips that might also be used to bash the lord's head in if it came to it.
Also worth noting, a lot of ISO certification is ridiculously easy to get. 27001 you can basically copy off some qms procedures to your google drive and call it a day
Why would I use Claude or ChatGPT when other companies (Meta, Alibaba, Forest Labs, Stability AI) are doing the same, but also giving you access to the base model for free?
Being able to run offline AI, or at least being able to host that by yourself gives you lot more freedom and transparency than a stupid ISO certification created by bureaucrats to comply with their own policies.
I have just disbandoned anthropic. I was trying to extract knowledge from some PDFs with academic papers about financial institutions.
It refused because of the content filter. And the recommended solution would be chunking.
I simply can not be prepared to handholding a LLM like a mad toddler for doing tasks like this.
The standard itself was created by Europeans, Anthropic itself is European, so it's just europe trying to impose their own vision of AI on everyone else. That's another example of thinking from perspective of the old world. The New World is busy building instead of putting regulations around the progress.
If you think like a citizen then it's better to avoid sharing your data with other AI companies and only share data with US.
Uh, no it's not? ISO is the international standards body whose members are the national standards bodies of various countries.
> The standard itself was created by Europeans,
So I drilled into this spec a bit, it comes from JTC 1/SC 42 [1], and the chair of that subcommittee seems to be ANSI, which is the US standards body. I don't know much about this committee (I'm mostly limited in knowledge to JTC 1/SC 22, and even then, the relevant working groups for the languages I work on), but their main website seems to be at [2], not that I can parse which WG actually worked on the document or the main working body. But I don't see many indications that this is a Europeans-trying-to-beat-down-Americans standardization process.
> The International Organization for Standardization (ISO) is a non-governmental international organization with its headquarters in Geneva, Switzerland.
Their employees are mostly Swiss, they subconsciously will want to protect their interest.
Having actually worked on standards produced by ISO, by the employees of ISO themselves do almost nothing on the standard. The standardization process is driven almost entirely by the members of the working group, with the main role that ISO itself plays is charging money for the final product and giving it an official ISO number.
Trying to claim otherwise just highlights your complete ignorance of the standardization process.
"Secretariat: ANSI (United States)
Committee Manager: Ms Heather Benko
Chairperson (until end 2027): Mr Wael William Diab
ISO Technical Programme Manager [TPM]: Ms Jacquelyn MacCoon
ISO Editorial Manager [EM]: Ms Jessica Navarria
"
well done, another US centric "international standard" with US rivals excluded from the very beginning. probably for national security reasons again?
Love or hate it, generative AI must be scrutinized for copyright infringement.
And it is coming. I don't have the slightest doubt about it.
The problem is not only hallucination, is mainly the data set.
If things are done right, there must be a technological solution for authors which work is used for training of the models to receive monetization.
One of the logical ideas is of revenue sharing.
There is no way everything inside popular generators to be fair use.
By the same argument, if stealing is illegal then that just means a few greedy people will have all the money.
There is a kind of techie trope where the logical thinking of computer science is applied to the law in a more creative way. And VC start up culture cranks this up to the next level by treating illegal activity as more than acceptable if the fines do not overly impact the profit margin.
There is definitely a subjective call for our society on what we want to do with copyright. I've been a proponent of copyleft for a long time but I'm not going to imply that the law as it stands condones piracy or plagiarism.
If this thing is not regulated for copyright, you know that communism is here.
There is no better analogy than power of the collective knowledge used by nomenclature and oligarchy. Some people know the suffering and horrors resulting from this type of societal contract.
Looking at AI as only a technology is wrong.
Technology has implementation. This makes the difference.
Megacorporations are investing in AI led by bankers into a future against human qualities and substance. This is clear but requires non-biased view, critical thinking and following the money.
ML can be applied properly and with respect for copyright and humanity.
The current crop of stochastic parrots is an insult to common sense.
I don't expect people which make money with this tech to be objective.
They love the power of the moment and the promising future of domination.
But let's be honest here, this is the heist of the century.
Marc Andreesen said in an interview he was told to not invest in AI startups by people in government because the government was going to decide which companies were allowed to research and develop AI and nobody else. Does this certification make Anthropic one of the chosen, golden child, ordained companies?
That can mean a lot of things. Someone suggesting it because they think that's where things are going versus someone warning or threatening. With the little I know of the man, I take anything he says about the government with a truck load of salt.
Yes I saw that too. Andreessen and Horowitz both believed that the last 4 years under Biden has been the worst environment for startups of their working life.
The Dems basically wanted to anoint a small number of companies at the frontier of AI and crypto space that they can control and people outside that were being "de-banked" etc.
Anything related to reverse engineering? Refused.
Anything outside their company values? Refused.
Anything that has the word proprietary in it? Refused.
Anything that sounds like a jailbreak, but isn't? Refused.
Even asking how to find a port that you forgot in the range between 30000 and 40000 with netcat command... Refused.
Then there's openai 4o that makes jokes about slaves and honestly, if the alternative is anthropic then openai can might as well tell people how to build a nuke.