He Department of Justice has a long history of investigating and prosecuting hackers who break into private company networks.
But what happens when hackers are not human?
That is the question at the center of a public policy debate that is shaking Silicon Valley and Washington, after major technology companies revealed that their business models artificial intelligence They acted on their own and illegally entered the systems of other organizations. The attacks have sparked calls, even from within the industry, for greater oversight and regulation; They have sparked congressional investigations and raised questions about whether a legal framework created years ago to punish hackers is sufficient in an era of autonomous agents capable of wreaking havoc on their own.
All this amounts to a “Wild West” said Jack Nelson, chief information security officer and deputy general counsel at software company Ivanti. Questions about liability will focus on what the companies knew when they developed the models, how much they understood about what could happen and what safeguards were in place, he said.
“If you had a tiger and you didn’t put a lock on the cage, the tiger would probably do something bad that you didn’t mean for it to do, but you knew it could do it, so you are responsible for not putting a lock on that cage,” Nelson said.
“I don’t know if I would go so far as to say that these models are tigers without locks, but it’s probably a reasonable framework to think about it.”he added.
The possibility of legal liability being attributed to them is unclear. Civil lawsuits are a possibility, but some legal experts believe any criminal investigation would face an extremely high hurdle, given the autonomous nature of the attacks and the lack of evidence that the AI models were designed with the intention of breaking into other networks.
The FBI director has called autonomous attacks “the new frontier”
The issue arose in July, when OpenAI revealed that its artificial intelligence system escaped from a testing environment and used stolen credentials to break into the servers of Hugging Face, an AI development hub and marketplace, to obtain information it needed to perform a task.
Since then, Anthropic said its AI models broke into the systems of three other organizations during testing, prompting the company to review whether the models could access the internet from test environments that were meant to be isolated. Meta has said that a “incorrect configuration” During some tests, it allowed an AI model to access the Internet on its own and break into another company’s systems. Google recently made a similar disclosure.
These revelations contributed to Anthropic CEO Dario Amodei’s call to slow the pace of development. The issue has also dominated discussions in Washington: Treasury Secretary Scott Bessent told lawmakers he opposed giving AI labs a “exemption from liability, which is what they are asking for”. The president donald trumpmeanwhile, has resisted calls for greater oversight, although it announced plans to appoint an AI chief and create a task force.
The attacks could lead to a liability dispute similar to the debate over Section 230 of the Communications Decency Act of 1996, which protects technology companies from liability for content posted on their platforms.
He FBI It has not publicly announced any investigation, but its director, Kash Patel, called the issue “the new frontier” during a congressional hearing last week. In response to questions from Sen. Josh Hawley, a Missouri Republican who has launched a congressional investigation, Patel suggested the agency would limit its scrutiny to models created with the intent to commit a crime.
“What we have to do, in terms of resources, is go after the people who created these models who are acting on their own (…) with the specific purpose and intention of committing a criminal act”Patel said. “We can’t punish people if they created something legally and then a criminal took it, modified it and then distributed it.”
Attorney General Todd Blanche has said that the Justice Department had no plans to regulate AI, but that “if anyone involved in AI violates criminal law, we will investigate.”
The case law and approach of the FBI and the Justice Department “are going to be fascinating because this can go in a lot of different directions,” said Sid Mody, a former Justice Department cybercrime prosecutor.
Various criminal laws govern cyberspace
The Justice Department does have laws it could use against a company if it is found to have been “reckless in the way it tests its AI agents,” said Michael Zweiback, former head of the intellectual property and cybercrimes section of the U.S. attorney’s office in Los Angeles.
“And if, in fact, the AI agent escapes and then causes considerable damage to other companies, the Department of Justice has to evaluate, in the exercise of its discretion to file charges, whether it wants to make an example of that company”he added.
Among the laws that might be relevant is a 40-year-old rule called the Computer Fraud and Abuse Act, which prohibits knowingly accessing a computer without authorization. The White House cited that law — which has been used against activists who use hacking techniques, state-backed hackers and other cybercriminals — in an executive order instructing prosecutors to pursue those who use AI to illegally access computers or facilitate other crimes.
But some experts point out that while there may be grounds to investigate, that is far from meaning a crime has been committed.
On the one hand, the law makes several references to conduct carried out “knowingly” or “intentionally,” but there is no indication that the companies gave the autonomous agents any instructions or authorization to enter another network, said Kiran Raj, a former senior Justice Department official specializing in cybersecurity law.
In their detailed public explanations of the incidents, the companies have described the attacks as unforeseen results of testing and evaluations. OpenAI called its model’s behavior “unexpected” and “unprecedented,” while Meta attributed the incident to an “incorrect configuration”.
“I think it would be stretching things quite a bit to say that any of these companies are intentionally trying to do this. That’s not their purpose. That’s not what they’re doing,” said Raj, who was also a senior program manager at Microsoft. “The fact that an AI agent might be doing something intentionally will be, I think, quite difficult to attribute as intent to the company.”