If a human hacker was caught doing close, they'd be inside for years, & AFK.
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
@JulianOliver Regardless of whether it was their explicit intention or not, a significant signal sent here is the effectiveness of these systems at laundering responsibility.
They successfully showed (once again) that you can do illegal shit and get away with it as long as you claim that the machine did it. Specifically, that their machine did it.
This is a sales demo.
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
@JulianOliver @blogdiva Been saying it since the huggingface story broke: Those assholes need to be prosecuted under CFAA. With the Morris Worm verdict, there’s precedent. Lock em up for a couple of decades.
-
P pelle@veganism.social shared this topic
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
@JulianOliver It's because they're allowed to pretend it's all "autonomous actions" with no causal link to their own actions, instead of the reality that they willfully and specifically programmed their computers to do these things (and then proceeded to not properly sandbox or supervise the processes). You don't get this kind of text prediction without deliberate model calibration after the initial "ingest the entire web" round (or perhaps also some corpus curation for the first rounds).
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
@JulianOliver@mastodon.social lock them all up. They know exactly what they’re doing.
-
@JulianOliver Regardless of whether it was their explicit intention or not, a significant signal sent here is the effectiveness of these systems at laundering responsibility.
They successfully showed (once again) that you can do illegal shit and get away with it as long as you claim that the machine did it. Specifically, that their machine did it.
This is a sales demo.
@chrisg @JulianOliver exactly. no agent woke up and did that on its own, or in response to someone's prompt for a powerpoint illustration
-
@JulianOliver Regardless of whether it was their explicit intention or not, a significant signal sent here is the effectiveness of these systems at laundering responsibility.
They successfully showed (once again) that you can do illegal shit and get away with it as long as you claim that the machine did it. Specifically, that their machine did it.
This is a sales demo.
@chrisg @JulianOliver that was my first thought too
-
These corporations get to enjoy the legal protections of a 'juridical person', yet when they mess with the rights of actual people, they don't suffer the accountabilities and due process we people do. The same old vampiric & classist exceptionalism, but now with trollface.
@JulianOliver we could fix that...
-
These corporations get to enjoy the legal protections of a 'juridical person', yet when they mess with the rights of actual people, they don't suffer the accountabilities and due process we people do. The same old vampiric & classist exceptionalism, but now with trollface.
Yup, the idea of Limited Liability as a necessary evil on a pragmatic basis has morphed into Corporations are people but with more rights than normal folk. We must all bow down to the mythical shareholder value as the ultimate "good".
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
-
J jwcph@helvede.net shared this topic
-
If a human hacker was caught doing close, they'd be inside for years, & AFK. Meanwhile, OpenAI gets to make everyone's infra their sandpit, & when damage is done, ride a 'look at the monster we made' investor gush.
"The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”.
OpenAI then confirmed it.
https://www.theguardian.com/technology/2026/sep/11/openai-agents-rubygems-malicious-packages
Seems like these days there are two option options for the criminally minded.
Go into government or start an AI company
-
@scott Indeed, a good point and perhaps even basis for a civil lawsuit, and/or some regulation with actual teeth.
@JulianOliver @scott maybe time for a "counter-terrorism" unit?...
-
These corporations get to enjoy the legal protections of a 'juridical person', yet when they mess with the rights of actual people, they don't suffer the accountabilities and due process we people do. The same old vampiric & classist exceptionalism, but now with trollface.
AI gone rogue is the digital equivalent of arson. You set the fire, you should pay for what is burned, wherever it is burned.
These companies need to be punished and contained when they do harm. Just as we would any roaming, violent psychopath in the population.
️
