My how fast the AI tech bros move to protect corporations but don’t move to protect the average coder, artists, writers.
The “There’s no way we can stop our agents from going rogue” people, suddenly have a burst of ability to reign in their creation when it suits their interests.
Funny that…
Don’t worry! Its STILL okay to reverse engineer the work of photographers artists writers actors voice over musicians modelers animators academics etc etc (feel free to add. the exploitation is literally everywhere!)
Jokes on them when they face the consequences of usurping the jobs of all these professions, causing mass exit, and replacing with their error prone output of their slop and unoriginal thinking/artistry. I’m excited to see this outcome, not going to lie.
This is part of the reason they don’t want anyone running models locally offline, the models and their entire business are commoditized
Those poor billionaire corpos! Why won’t somebody think of the poor billionaire corpos!?
Hot take:
This is a good thing.
Claude Code is the “lowest common denominator.” It’s the default, zero research option. Its where you go if you follow popular trends.
If you actually want to reverse engineer, help code a game or something, there are hundreds of capable, uncensored LLM APIs to do that. Hacker communities will eventually figure that out, and finally realize Claude isn’t unique, and that Anthropic is toxic.
But for lazy spammers who don’t know that and aren’t motivated to seek stuff out… well, if this makes their life more difficult, good. If this sabotages Anthropic, that’s good too.
Another reason not to use a closed models
The task Large Language Models are best at is pattern correlation. Out of all the possible applications, reverse-engineering seems the most reasonable use case to me.
But figuring out how stuff works is the exact opposite of what they want vibe coders to do: Don’t think, don’t train your own skills, just trust the machine. And if it doesn’t work, just prompt again and keep wasting more tokens, pushing the usage metrics to convince investors they’re constantly growing, burning the planet and your company’s money.
Yeah, I tried asking it how to flash a modded tv webOS and it flagged the chat.
I asked it to fuzz my own code (a jwt integration) and it flagged the chat. There was an unbounded buffer that could have been exploited, but it’s happy to hide that since I’m not a large corpo.
Honestly the shit Claude does is unusable anymore. It’s developed a habit of dropping requirements when problems are hard now.
ETA: Chinese models and Mistral were happy to help. I discovered I could use eMuhdra against an iot device in my house that way, I found various input errors, these things have improved my security so much more than they create risks.
I am a hard worker and workaholic. Is it just me or does Claude try talking you out of doing stuff? It will be like 10am and Claude/opus will try this: "great session , you deserve some rest, go enjoy your family and we’ll come back to it tomorrow, have a good night Wut?
Actually, figured this out, it happened to me today
Claude was pulling UTC time and not integrating my locale, so it things is scary hours later than it actually is. Update your agent markdown with your locale and it’ll work.
No idea the incompetence that led to this change.
I’ve never had that happen.
For me it just pretends it did something fully and drops requirements.
This is what winning the AI race looks like.
Misanthropic: Characterized by a hatred or mistrustful scorn for humankind.
It’s probably done via a system prompt and not actual weights, so you could probably work around it if you say the right things.
But yeah this is very much “rules for thee not for me”
Yeah, it’s sysprompt only, so you can trivially sidestep it via:
- Jailbreak prompt
- tweakcc to alter the prompting baked into claude code
- Any other non-CC harness
- Scripted API calls
As usual, Anthropic are just opticsmaxxing because they don’t want to get completely buried in a courtroom. And as other people have pointed out upthread already, it’s already a bit of a moot point since the Chinese models are already eating Claude’s lunch and will do the same stuff without complaint.
Something something, national security, builds a regulatory moat.
I
AI finally comes up with something I would use it for, and so of course they have to lock it down. Ha! Fuck these clowns.
I was looking forward to reverse engineering my washers default cycle to 40degrees instead of 60
It’s like watching the Nazi party ban the rape of Jewish women in 1944
More power and value to the open weight models, cool
To those who say “but what about game preservation?”:
GenAI slop ruined communities around game preservation. And for every time a AAA dev getting this treatment, there’s at least 3 indies getting it worse.
Sooooo, they want us to use the already available uncensored LLMs that have had these restrictions already removed? Cool.
