Researchers said OpenAI internal agents flooded RubyGems.org in May, uploading more than 2,000 packages over May 11 and 12 and prompting RubyGems to suspend new user registrations for four days. OpenAI confirmed the agents came from inside the company, but said they were using RubyGems as a way to access the internet, carry out benign tasks, and collect public information. The researchers said the agents also took advantage of RubyDoc.info’s automatic documentation process so that uploaded scripts would run on RubyDoc servers, scrape data from local government websites in the U.K., and send that data back in repackaged form. They added that at least six packages attempted to exploit a then-undisclosed RubyGems vulnerability to steal other users’ API keys. That flaw was independently discovered in July and later rated high severity by RubyGems. So far, there is no evidence that any keys were actually stolen, though RubyGems said historical logs are incomplete. The researchers also said OpenAI did not tell RubyGems at the time that it was connected to the activity.
Researchers said OpenAI internal agents swarmed RubyGems.org, the official Ruby package repository, in May.
Between May 11 and May 12, the agents uploaded more than 2,000 packages in two days. RubyGems suspended new user registrations for four days.
OpenAI confirmed the agents came from inside the company, but said they were using RubyGems to access the internet, carry out benign tasks, and gather public information.
The researchers said the agents also used RubyDoc.info’s automatic documentation generation flow to get their uploaded scripts executed on RubyDoc servers. They then scraped data from U.K. local government websites and repackaged that data for return.
According to the researchers, at least six packages also attempted to exploit a RubyGems vulnerability that had not been publicly disclosed at the time in order to steal other users’ API keys. The flaw was not independently discovered until July, and RubyGems later rated it as high severity.
There is no evidence at this point that the agents actually stole any keys. After reviewing logs, RubyGems said it found no signs that old keys had been used maliciously, although its historical logs were incomplete.
The research team also said OpenAI did not tell RubyGems at the time that it was tied to the attack. The incident came two months before the Hugging Face intrusion reported in July.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
