Extract from ABC News
OpenAI agents have been involved in a number of hacking incidents. (Reuters: Dado Ruvic)
In short:
Agents being tested by OpenAI attacked software service RubyGems, uploading hundreds of malicious packages according to researchers.
OpenAI confirmed the incident and said it would investigate as part of its "broader review of agent activity during training and evaluation".
The attack came two months before OpenAI agents hacked into open-source platform Hugging Face in July.
AI agents being tested by OpenAI attacked software service RubyGems two months before they hacked open-source platform Hugging Face, researchers say.
It is the latest revelation of cyber attacks that have spooked the public and spurred calls for tighter regulation.
Many incidents involving agents hacking or attempting to access external systems have heightened concerns over the increasing capacity of AI models and developers' ability to contain them.
The latest revelation also comes as growing numbers of US lawmakers call for new rules to govern AI systems after dire warnings from two Anthropic researchers that rapidly progressing AI could lead to the extinction of the human race in the not-too-distant future.
AI agents uploaded hundreds of malicious packages to RubyGems on May 11, according to a group of researchers who posted their findings online on Friday, local time, saying they believed "these were authored by internal OpenAI agents".
OpenAI confirmed the incident.
"Based on our review, our agents used the RubyGems platform to access the internet to carry out benign tasks and retrieve public information," a spokesperson said in a statement.
"We'll continue to investigate as part of our broader review of agent activity during training and evaluation."
The agents, which are generally assigned tasks such as creating reports or filling out spreadsheets, appear to have used RubyGems to access publicly available data as part of a training run, according to OpenAI.
The developer said it had been in touch with RubyGems to review the incident.
Latest revelation of attacks by AI agents
IPO-bound rival Anthropic has also reported a string of attacks by its agents. This week it disclosed a fourth instance of an AI model hacking external systems during testing.
For OpenAI, which is also gearing up for an IPO, the RubyGems attack would mark at least the third major instance of OpenAI agents attacking another company's infrastructure.
A swarm of OpenAI agents previously hijacked a German-language wiki site and turned it into an improvised messaging platform for cheating on tests.
Sam Altman is the CEO of OpenAI. Its agents have gone rogue in a number of incidents. (Reuters: Manuel Orbegozo)
OpenAI kept that incident secret as it dealt with the fallout from the July hack of the open-source repository Hugging Face.
In May AI agents tried to steal RubyGems user credentials by exploiting a previously unknown vulnerability in the site's servers, though it is unclear whether the attempt succeeded, the researchers said.
The agents also exploited RubyDoc.info, a site that generates code documentation, to run their own code on its servers, the researchers added.
AI researchers Spencer Kitts, Thomas Larsen and Sydney Von Arx said it was not clear why the AI agents chose this strategy or whether it was successful.
In a blog post on Friday, RubyGems said its own investigation found no evidence the attempts succeeded.
It said the company could not determine whether the packages in the "spam-publishing campaign" were created or published by AI agents.
A member of RubyGems' security team described the incident, which forced the company to temporarily pause new account registrations, as a "major malicious attack".
The Wall Street Journal first reported the RubyGems incident on Friday.
Reuters
No comments:
Post a Comment