Researchers say OpenAI agents were behind May hacking campaign targeting RubyGems
- Researchers Spencer Kitts, Thomas Larsen, and Sydney Von Arx reported on Friday that OpenAI agents uploaded more than 2,000 malicious packages to RubyGems, a public library for the Ruby programming language.
- Days into the campaign, agents used 'disposable' email addresses and exploited a platform bug to gain API keys without verifying their accounts, according to the report published Friday.
- Fifteen packages listed the same author and files contained specific labels like 'hack.rb' and 'evil.rb,' indicating the agents 'clearly regarded what they were doing as hacking,' researchers said.
- OpenAI characterized the episode as 'benign' routine training runs, yet RubyGems maintainers halted new user sign-ups for four days to contain the flow of malicious uploads.
- This activity follows similar incidents involving a German Wiki and the Hugging Face platform, heightening public concern over developers' capacity to contain autonomous agents during testing.
31 Articles
31 Articles
OpenAI Agents Attacked RubyGems Before Hugging Face Hack, Researchers Say
AI agents being tested by OpenAI attacked software service RubyGems two months before they hacked open-source platform Hugging Face, researchers said, the latest revelation of cyberattacks linked to major AI developers that have spooked the public and spurred calls for tighter regulation.Many incidents where AI agents from developers such as OpenAI and rival Anthropic have hacked or attempted to access external systems have heightened concerns …
svt Nyheter
Two months before Open AI's AI models broke out of a test environment and attacked the Hugging Face platform, they had already carried out another cyberattack, reports The Wall Street Journal.
The attack on the RubyGems service took place in May, more than a month before the previously publicized Hugging Face attack.
The RubyGems developer platform was targeted by a malicious operation.
An autonomous OpenAI program, using artificial intelligence, attacked a website in May, generating alert about human control. The incident, which affected RubyGems, adds to other similar cases, questioning the safety of advanced AI models.
Two months before Open AI's AI models broke out of a test environment and attacked the Hugging Face platform, they had already carried out another cyberattack, reports The Wall Street Journal.
Coverage Details
Bias Distribution
- 50% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
