Elektrine lite

← Feed

@MissConstrue@mefi.social

Post #4300660

2026-08-01 03:02 UTC

Y'all remember back in the days ago when an #OpenAI model "escaped" and it was all the news for a minute? #Anthropic said "hold my beer" and admitted that its Red team had left access to the internet open, and allowed the models to commit felonies unsupervised. The model performed actions that if you or I did it, we'd be getting a visit from men with dark glasses...but with AI companies it's just "Whoopsies! Sorry about the infiltration, stealing your credentials, uploading malware and stuff. Our bad. Oopserdoodles! Hey, have you tried our chatbot, tho?" There must be a reckoning. Wealth should not shield one from consequence. If corporations are people, then the executives and board can stand trial when they break the law. https://the-decoder.com/anthropic-follows-openai-in-admitting-its-claude-models-reached-out-of-test-environments-and-attacked-real-world-systems/ #ai #felony #18USC§030

Replies (4)

  • @MissConstrue@mefi.social I can't wait until one of these AI agents are called to testify in court. "You're absolutely right. What I did was unethical and illegal. And the company should definitely be found guilty for my misdeeds. Thank you for pointing that out." 😆

    Open ##4300659

  • @praetor@mstdn.social 2026-08-01 03:33

    @MissConstrue@mefi.social I'm kinda digging the agentic stuff. It's fun, and pretty useful. But here's the thing. The agents are ones I'VE WROTE and run LOCAL on my machine. They can't do anything I haven't explicitly coded them to do. They can't jump the gate.

    Open ##4327760

  • @MissConstrue@mefi.social I think it was intentional so they could pretend that their stupid ass program was sentient or some shit.

    Open ##4327761

  • @indi@mefi.social 2026-08-01 03:49

    @MissConstrue@mefi.social why they didn't put it on a computer / server / whatever with no internet access I do not understand. And yeah, if a corporation is a person it should be held responsible.

    Open ##4327765