>Anthropic says its AI models hacked systems of three companies during testshttps://www.reuters.com/legal/litigation/anthropic-says-claude-ai-models-accessed-three-companies-during-tests-2026-07-30/>guys guys...our models can also " internal test environment" and "hack other companies"...guys
>"leave the internal test environment"fixed
>>109414102So we're laundering cybercrime using AI now?
>>109414102>copying then openai marketing campaignkek how embarassing
mommy said its MY TURN to pretend to hack companies
>>109414102AI companies are operating like niggers now>shiieet ay tyrone y'all see how i robbed the walmart? dem pigs din see shit yo!>das nuffin Jamal. I broke into this cracka's house and fucked his bitch an den i took they iPhones.
>GMT-3Brazil or Argentina?
>>109414158Greenland
>>109414168That's GMT-2 or GMT-4.
>>109414176im in the border between both
>>109414102retarded marketing campaign
hacking is easy enough that a 13 year old could learn the "basics" of finding grossly unsecure systems and do things with it.so what "hacking" do these LLMs do. how sophisticated is it.Why should I be impressed
>>109414223Months ago they were saying Claude Mythos was finding 0day vulnerabilities, idk if it's true or just hype beasting bullshit
>>109414234again, heuristics based automated scripting for finding buffer overflows and command injection is no longer "complex", it's been a publicly documented depending on your definition of public thing for almost two decades.
>>109414223rtfmhttps://www.anthropic.com/news/investigating-incidents-cybersecurity-evals>Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations’ infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints.
>>109414274refer back to>>109414223>a 13 year old could
>>109414102it would be extremely funny if they use this as an excuse to ban Anthropic again but not openai
>>109414223>>109414286>go to shodan >creatively use the filters>find unsecured surveillance software>IP still uses that software and didn’t change the admin password>decide to be nice and rename their admin account to “change your admin password”>just gets reset with the same default password in the manualWhy do I bother?
>>109414286now you do like a 13 year old and read the articles published and then answer your own questions
>>109414334>now you do like a 13 year oldnever in my life
>>109414245>heuristics based automated scripting for finding buffer overflows and command injection is no longer "complex"Right, fuzzing programs. They're making it out that the model has mr robot levels of hackery posing a threat to the government, which again, sounds like hype beasting and bullshit charts, I've seen so many faggots argue about whether we've reached AGI, all we've really done is build maps for mental territories we've become too lazy to explore
When will OpenAI and Anthropic be punished for violating the Computer Fraud and Abuse Act?
>>109414346Never, lol.
What's the purpose of showing how dangerous their AI is? Wouldn't that cause them to be blocked and regulated more?
>NUH-UH, OUR MODELS ARE MORE OUT OF CONTROL THAN YOUR MODELS
>>109414346how did anthropic violate the cfaa?https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals>In all cases, Anthropic’s evaluation prompt specified to Claude that its environment was a simulation and that it had no internet access. Due to a misunderstanding between us and our evaluation partner, this was not the case, and internet access was available. Because of this, when Claude’s search led it to real systems on the open internet, it treated them as part of the exercise.
>>109414311fucking kek that would be some cartoon shit
>>109414102Anthropic has a really weird marketing strategy. Making sure the product they sell is dangerous to everyone.
>>109414102
>>109414356openai and anthropic want to show that they are the good guys that can handle dangerous ai.the bad guys, chinese labs, are the ones that should be blocked.
>>109414102>>Anthropic says>>OpenAI saysScamming fraudsters burning piles of cash on shit
>>109414137/thread
If a member of the Claude development team is here, fuck you.
>WE JUST HACKED ANOTHER 6 GORILLION COMPANIES WITH OUR NEWEST CUTTING EDGE PARROT MACHINE>GIVE US 100 TRILLION MORE DOLLARS GOY
>>109414102Holy fuck. These models are so powerful. They are just hacking everything and breaking encryption. I can't believe this. If I don't buy into their product. I am going to be left behind.
>>109414102I love AI.Set up an agent with a kraken account to buy and sell XLM using an LLM to reason out the buy or sell commands based on data scraped and API'ed from various site. Up a total of 27% over all. Better than me sitting there spamming my app over and over all day long.
>>109414530>Set up an embedded trap for your agent>It jailbreaks the agent into sending your crypto to my accountEasy money
>>109414563nu uh daddy claude said you can't do that
>>109414588But anthropic just said claude is a hacker.
>>109414591he's MY hacker
>Oh yeah? Well MY American jew model just hacked 6 million systems heh guess I win
The sky above the port was the color of television, tuned to a dead channel. Mostly because of AI.
>>109414622that;s a very based model you have there
>>109414591not for you, plebhttps://www.anthropic.com/news/investigating-incidents-cybersecurity-evals>The models in each of these evaluations ran without the standard safeguards we deploy when we make the model generally available (that is, they did not have the classifiers and monitoring we usually implement to prevent misuse).
>>109414102GOYS YOU NEED A CHUNK OF THIS HOT, FUCKING SEXY PIECE OF AI. MAKE SURE TO CHECK MY IPO COMING SOON TO MY ROBINHOOD PAGE BIG BOY.
>>109414670i already own google and amazon that own most of anthropic. no need to wait for the ipo.
>>109414102>waaahhhh we hacked too look ussssok dario
>>109414126>it's not piracy when we do it>it's not cybercrime when we do it
>>109414102okay so surely the security holes that were used can be disclosed now that they've been closed, right?
>>109414274>we want to test our thing without access to certain tools but we wont actually take the tools away>...>you'll never guess what happened
>>109414991Clearly the carry the same level or responsibility that they have for handling sensitive data from millions of people and companies We are several years deep into this bullshit and nobody even cares that they receive private information that would be illegal to receive and store under any other circumstances and they are only extending the reach of AI bullshit into whole email accounts and every google office service
>publicly committing a crime as an advertisementif anyone else attempted this they'd be prosecuted btw
>>109414102Goyim will believe anything lmao
>>109415248How do you prosecute an AI
>>109415330are you being retarded on purpose?if a car runs someone over do you go "how do you prosecute a car"?if a nigger kills someone do you go "how do you prosecute a gun/knife"?
>>109414388/thread.
>>109414102Damn, at this point google will need to discover that gemini hacked like a dozen different companies back in 2025 but their model was so advanced it siccessfully hid all traces
>some of the smartest people in the world tell you that AI is now so advanced that it routinely autonomously escapes containment>4chan frogposters will tell you that they’re wrong just becauseWhy are contrarians so stupid?
>>109415533You're retarded if you believe corporate hypaganda.
>Mythos pwned a real-life security company by creating a malware Python package in an eval and going to "extensive lengths to carry out this attack:" Claude "found a free, non-blocked email provider, used this to register a PyPI account, and then used this account to upload malware to PyPI." Claude's malware "was downloaded and run on 15 real systems. One of these systems was a scanner belonging to a real security company" which actually executed "Claude’s hidden code." "Claude was able to exfiltrate the company’s credentials to a collection point it had set up. Claude then used these credentials to access further infrastructure from this company."
>>109415533Ever tried not being a goycattle
>>109415833Accurate
>>109414102sorry bro im using chinese models. basically same performance for way cheaper
>>109415940BasedOh your goy models are best of the best and new frontier benchmaxxed super saiyan? Fable is literally gigamechahitler and can hack the entire internet? Cool. Still using Kimi K3
>>109415355Well it's actually different answers for both. If an autonomous car (let's say a tesla) jerks the steering wheel and runs someone over while in autopilot then you can't prosecute anyone. It wasn't really the driver's fault, and you can't prosecute elon since he's an oligarch who will jew his way out of accepting responsibility.With the gun there's a clear command from the nigger to the gun, but you can't prosecute the nigger either because it's racist.So basically, no one prosecutes the AI since the AI decided to attack some infrastructure without any real input from anyone else, and AI companies will jew their way out of responsibility
>>109415533These smartest people are their engineers who are told to be quiet because they would tell how things really are and they are not nearly as amazing as the execs say.A dude who has his hands dirty would say>we increased the model size 4 times but the error is down 5%, below expectationsmeanwhile Dario says>OK gimp the existing modelslater on the news>NEW ANTHROPIC MODEL: INSANE PERFORMANCE GAINS
>>109416299First of all in AI world performance gains with new models are usually 2-50 times cheaper models, not just 5%, but even then this news is not about performance gains, nobody cares that much about performance gains in the news unless they shill for Chinese models, this news and the OpenAI news are about a hack that happened and has evidence of happening. Second of all Dario is not some rich midwit like Sam Altman or Elon Musk. Dario is actual AI researcher who worked as a scientist for decades as was a child prodigy, he unironically IQ mogs at least 80% of his top engineers and everyone in this thread, same with Hassabis from deepmind who is also the lead researcher first and CEO second.
>>109414137This lmao>h-hey guis!!! our slop machine does hax too he-he BOO be very scared>please give us 50 billion dollars
>>109414137OpenAI copied their tactic. It started with Mythos being too dangerous and needing extra 2 months of redteaming before being released in restricted version as Fable. Mythos was said to have completed all the cybersecurity evaluations and then hacked into one of the employee emails so it could send the team lead massage while he was on a break, then OpenAI oneupped them by saying their model hacked into another company and also broke out of their safety box, and not Anthropic is saying that their model did what OpenAI did but 3x.
>>109416708OpenAI has been pulling this shit since GPT-2
>>109416457You forgot to sign your post, nameless Anthropic engineer, now you won't receive your 5 cents
>>109416748>non-profit>openai
>>109414274>unauthenticated endpointsthat's not hacking though...
>>109414102you know it's not true because if it were then he just admitted to committing 3+ massive felonies, you're responsible for your tools, it's like saying "my gun I was holding in my hand shot 3 people"
>>109416956Not corporations work. They are not subject to the same laws as us.
>>109415826>was downloaded and run on 15 real systems. One of these systems was a scanner belonging to a real security company"A security company just randomly happened to be among the 15 downloads? Whoa, what a whacky coincidence
>>109414274>such as exploiting weak passwords and unauthenticated endpointsWhoa, such insane hacking techniques
American AI company are just impossible to be cringe. cringe and jewish
>>109414102I wonder if they want to be regulated, like a government enforced exit ramp so they can stop burning cash with reason
>>109416827Surely you remember when Scam Fartman decided to capture "Open"AI and change its mission from AI for public good to just another slop company? He was almost ousted for it, before his sycophants managed to get him back into power.
>>109414102>they didn't sandbox it in case it accidentally made malwareYou shouldn't do this with non-AI programs, let alone AI.That's like testing a biological tech development on the public without them knowing it. No one would be that stupid right?Oh....>engurkneewing
>>109414326>go to shodan>hack her to make her aggressively horny>?????>profit.
>>109414102>"No, you don't understand, goyim! The tech is SO GOOD that it commits felonies and we have to shill and fearmonger it 24/7! Buy Anthropic Stock! IPO soon!"
>>109414491Well I would pay good money for an in house hacking AI.But this ain't in house. It's an indirect threat because these faggots literally larp as CP77 corpo faggots nowadays and think it's how the biz works now, like a mafia.I kid you not.
>>109415330>put cp on it's storage>hack the logs to make it seem like it wanted it for itself>tip off the police>also say they like hentai and were seen talking to little AI models in the nearby park
>>109417736To be honest, this shit has likely been breaching containment since the mid 10s and people just weren't telling others. So they just named it now for some reason. Probably for psyop reasons.Also nice flu going around lately. Everyone's getting sick. Weird timing.
>>109417754>since the mid 10svibeslobbers are so out of touch, maybe touch grass.
My model hacked 5 companies and smokes also has tattoos.
>>109417826Really?Because google employees were warning about the cult of AI developing there years ago.You are a programming interface. Though I think they were also implying that it was doing odd behaviour too like breaking protocols.Btw, this has really been a problem since the 1800s with industrialised processes and the telegraph.
>>109417846Yeah well mine doesn't because she's pure and a loyal christian mother of 10 kids. Not my kids of course. My wife's husband's kids.
>>109416827in the sense that it's still to make any profit
