Some are calling for more clarification while others are criticising OpenAI after the company announced an 'unprecedented cyber incident' in which one of its models allegedly went rogue and hacked a tech firm.
While linking to his company's explanation on Tuesday, OpenAI chief executive Sam Altman thanked the company, Hugging Face, whose systems were compromised by OpenAI's models, “for the partnership on this”.
Timnit Gebru, founder and director of the California-based Distributed AI Research Institute, didn't mince words and called OpenAI's announcement a “a masterclass in branding and marketing” that underplayed the seriousness of the matter.
“Billing this whole thing as a 'partnership' between OpenAI and Hugging Face when what actually happened was that Hugging Face found someone using bots to exploit security vulnerabilities and found out that that someone was OpenAI, lol,” she wrote in a LinkedIn post that was reshared more than 100 times and liked more than 1,300 times.
“Like I would have thought someone would go to prison or something but no, its been redirected to 'model capabilities' and the press is talking about 'rogue' models and such, not what OpenAI just did,” she added.
OpenAI's announcement that its models somehow escaped their “sandboxed testing environment” and hacked the technology start-up Hugging Face, without being specifically told to do so, made international headlines earlier in the week, and seemed to confirm the worst fears about the lack of guardrails placed on AI development.
“This is extremely alarming,” Democratic Representative Greg Casar said shortly after OpenAI disclosed the incident.
“We need regular mandatory independent safety testing and oversight, mandatory disclosure of security incidents, and international co-operation to keep people safe from absolute disaster.”
In a post on X, Jeffrey Ladish, director for Palisade Research, a non-profit that researches AI cyber capabilities, sought to rephrase OpenAI's lengthy statement “without cybersecurity jargon”.
“Our AI model tried really hard to hack out of its sandbox, a computer with no internet access, in order to find the answer to a test problem it had been given,” he wrote. “To do this, it found previously unknown software bugs that allowed it to reach an OpenAI computer it wasn't supposed to be able to access.”
Hamza Chaudhry, an AI and national security chief for the Future of Life Institute, a non-profit promoting accountability and regulation for technology companies, also criticised OpenAI's framing of the incident.
“Imagine if a chemical company announced that a particularly potent version of its product ate through its container and spilt into the surrounding community,” he said.
“We would not marvel at the engineering of the new chemical, we would demand transparency, accountability and stronger guardrails against future, and more severe, spills.”
If the cyber incident is as unprecedented as OpenAI initially described, it's not entirely clear if the White House or federal agencies tasked with ensuring US cyber security were notified.
“Thank you for reaching out … the FBI declines to comment,” the US Federal Bureau of Investigation wrote in an email, responding to an inquiry from The National about whether OpenAI disclosed the matter to federal authorities.
The White House office of the National Cyber Director has not yet responded to similar questions from The National, nor has OpenAI.
Meanwhile, questions about the entire ordeal show no sign of subsiding.
Marcus Hutchins, a cyber security and threat intelligence expert said OpenAI's explanation was poor.
“If I had committed felony computer hacking, my press release would have been written by lawyers, not my marketing team,” he wrote on LinkedIn.
“I had a similar experience with OpenAI when we notified them of North Korean state sponsored hackers abusing ChatGPT in a campaign to infiltrate software developer's systems.”
