A Ford driver has to choose to get behind the wheel of their automobile, drunk, and hit the nearest minivan. A human being is exercising intent; that human can be held responsible.

In this case,

> He said the incident began on June 18 when an OpenAI research team used an internal model to conduct internet-based research into the public medicine space.

> The AI agent encountered repeated blocks while seeking information from the government portal but found ways around them, ultimately gaining unauthorised access to other areas.

It sounds like the OpenAI team didn't ask for attacks/bypasses, the emergent behavior of the system decided the best way to satisfy the goal was to go rogue. Why shouldn't the manufacturer of a defective product be held accountable?

This wasn’t someone using OpenAI to break into a government/company, this was OpenAI using OpenAI’s models, and then breaking in. It was a research team that WORKED FOR OpenAI.

We don’t fault Ford for drunk drivers, but if the CEO of Ford got wasted and drove his car into another one we would definitely be charging him.

Yeah I mean if Ford execs run a team of workers that drive drunk for "research purposes" then yeah they should be rounded up.

> He said the incident began on June 18 when an OpenAI research team used an internal model to conduct internet-based research into the public medicine space.

I think he means hit them with criminal charges for breaching and hacking other companies unintentionally, not because they make a tool that could potentially do something like that. I still think it wouldn't be right, nor realistic in today's political climate in the USA that any criminal charges will come for unintentionally hacking sites.

This was OpenAI. It is as if Ford employees ran over people. They absolutely should be charged. We would be charged as well.

openai made the tool and also used it to commit the crime

Yes. Because then we'd have breathalyzer interlocks, speed governors, etc. auto makers were also held liable for other safety issues. It would seem that the RATE of such requirements has to be limited in some way, but not that they not exist at all.

You fucking know that’s not the same thing. Jesus fuck. Your AI Delusion Sydrome is showing.

OpenAI and Anthropic should absolutely be fucking held responsible when the thing they created AND control are hacking shit.

Get your head out of your fucking ass.