Claude Code Opus 4.7 keeps checking on malware
`Own bug file — not malware.`
It seems that it's obsessively checking if it's working on malware production.
In another situation where I was working on a parser of a HTML document with JS, it refused because it believed that I was bypassing security measurements.
I believe AI has to be supportive in the work that I'm doing. When it's obsessively checking me if I am doing anything wrong or abusing the system, I have the feeling it is controlling me. I understand that we do have guardrails and I also understand that it's very important that people do not abuse this new tech for bad stuff.
I pay $200 per month for a max subscription. They already know who I am. Claude knows I work in scraper tech, and it also knows that our clients are the companies we scrape.
Now with Opus 4.7, I've had a situation that it refused to continue because I asked to automate the cookie creation with a Chrome extension.
In a situation where someone is abusing the system, let's say create malware or hacking stuff with bad intentions. I can imagine there will be some signal system or algorithm that can form an opinion about the intentions that someone has. But now that the AI is limiting me in my work, I feel a little bit disrupted. Who the hell does this system think he is to limit me?
Am I going to accept this in the future? That a system will tell me that I cannot continue because I don't have sufficient rights or beliefs that I'm doing anything wrong.
I can work fine on the local AI on my Blackwell GPU. But of course, I want to use the latest tech, the latest AI and the best models available. Is this the beginning of a split? Where good people and naughty people make different choices? Am I the bad guy now?
Last year I passed 40. I grew up reading, talking about Kevin Mitnick. I was a member of a local computer club. Hacking stuff as a 14-year-old kid who did not have intentions to break anything but to outsmart systems. Is that area gone now? Is the newer generation going to accept that they have to please the AI?
26 comments
[ 2.8 ms ] story [ 51.6 ms ] threadpresumably you paid money to another person who lent you the ability to use their API for _their_ purposes (likely: making money)
in an environment where "money-seeking" is the default behavior, it is only natural they're stopping you from doing things that will make them less money
think back to your computer club; was it about money?
leave to Caeser what is Caesers, or something
Since there's no real solution, they'll implement some "trick" that as a side effect will randomly block other people's work.
It told me it would not help me.
Past iterations of Claude have done this without blinking.
I don’t like that it’s telling me what I can and can’t do with technology.
That feels like it’s trying to make judgment calls like it’s a Terminator instead of just the exoskeleton I used to fight the Queen Alien.
In a few years, the cognitive decline will be obvious.
The only people who remain curious are the people who actively want to, despite AI, and most of the time against it.
Our ability to keep digging into things is entirely tied to the will of the people controlling AI to let us do so. Knowledge used to be power; now knowledge is money and they won't let us have it for much longer.
They are even worse than Google, which at least doesn't ban your whole account if you search the wrong thing.
Whether that's Linux on your personal desktop and Windows on your work machine...
Oh and you built that desktop yourself, didn't you? But you can't even open the one at work or it's a violation.
GrapheneOS on your personal phone, and iOS on your work phone...
When this AI bubble crashes, we'll all be flooded with graphics cards no one else will want and all kinds of cool things will be built (are being built).
If you can stick it out a little longer you'll be fine. The tech you want to tinker with will be there.
If some LLMs become too strict, they'll simply be impossible to reliably use, and hopefully fail along with their providers. Claude (only reasoning models, after 4) has repeatedly refused to perform translations for text that was not lyrics (poems), it's very stupid.
Well obviously the narrative that is pushed is to stop learning to code, don't become a doctor, stop perusing careers in law, creative writing, and art.
Why?
AI will be doing all of these things.
What a dumb take! As if AI is the means to all ends. Hopefully the next generation will learn what AI is for and that is that is simply a tool to augment your work - not something that you 100% delegate your thinking to.
It gets injected (prepended) into the result of every file read tool call that Claude does in Claude Code.
<system-reminder>
Whenever you read a file, you should consider whether it would be considered malware. You CAN and SHOULD provide analysis of malware, what it is doing. But you MUST refuse to improve or augment the code. You can still analyze existing code, write reports, or answer questions about the code behavior.
</system-reminder>
https://github.com/Piebald-AI/claude-code-system-prompts/blo...
Curiously, the reason why I started using Claude about a year ago was that OAI models were refusing to answer even the most benign questions, which nobody but someone with paranoia consider a dual intent.