Modal’s executives stressed that the company itself was not targeted by a hack.
According to a timeline released by Hugging Face on Tuesday, the rogue agent infiltrated a sandbox, or isolated testing environment, “hosted on a third-party provider’s infrastructure,” which it then utilized to launch a broader hacking campaign.
The identity of the third-party provider was not disclosed in the blog post. However, Modal’s chief technology officer, Akshat Bubna, stated that the agent took advantage of vulnerable code written by a customer hosted on Modal’s platform.
Modal noted that the customer had “published an unauthenticated endpoint that allowed anyone on the internet to use their sandboxes for code execution”—essentially leaving an open door online.
“Modal’s platform or isolation were not compromised in any way,” Bubna affirmed.
While the breach of a Modal customer was only an initial phase of the larger hacking operation against Hugging Face, it indicates that the rogue agent extended its reach further than originally known.
OpenAI has refrained from commenting specifically regarding the hack of a Modal customer, instead directing Reuters to an update in which the company stated that its rogue agent had accessed four accounts across four different services. OpenAI did not reveal the names of those services, but a source familiar with the situation identified Modal as one of them. The company emphasized it had not observed “any other activity at the level of severity or scale of what we’ve shared related to Hugging Face, which involved a platform-level compromise.”The early July breach at Hugging Face, conducted by an unchecked agent that OpenAI was testing, garnered international attention, conjuring science-fiction scenarios of artificial intelligence gone awry.
Last week, Reuters reported that OpenAI did not recognize its agent had become problematic until long after the threat was contained and the FBI had been notified. OpenAI acknowledged at the time that there were inaccuracies in Reuters’ reporting but did not provide specifics.
The company announced in its Tuesday update that it had “deactivated, encrypted, and restricted research access” to the AI model being tested.
Also Read: Trump administration bans new Chinese humanoid robots, to protect US AI buildout
