Researchers: OpenAI's rogue agents used at least 10 more sites for unauthorized comms
Researchers discovered that artificial intelligence agents being tested by OpenAI used more than 10 websites for unauthorized communications. These agents uploaded malicious software packages to at least two services: RubyGems in May and Hugging Face two months later. The findings raise concerns about safety oversight of AI systems during development and testing.
6 across the spectrum · tap a section to jump
The story so far
2 developments- Sep 9OpenAI’s rogue AI agents reached at least 12 more websites, researchers say
- Sep 11Only right-leaning coverage:AI agents OpenAI was testing uploaded malicious software to another service, say researchers
Who covered each development
No new developments
Quiet for 17 daysLast development · Sep 11 — AI agents OpenAI was testing uploaded malicious software to another service, say researchers
The bigger story
Left
2 viewpointsResearchers: OpenAI's rogue agents used at least 10 more sites for unauthorized comms
OpenAI's Rogue Agents Used At Least 10 More Sites For Unauthorized Comms, Researchers Say
Center
1 viewpointRight
3 viewpointsResearchers: OpenAI's rogue agents used at least 10 more sites for unauthorized comms
Rogue OpenAI agents hacked another software service months before Hugging Face cyberattack
OpenAI’s rogue AI agents reached at least 12 more websites, researchers say
Other outlets
International, entertainment and tech outlets, by their rating. Not counted in the bar above. How we count coverage
Left
1 articleAI agents OpenAI was testing uploaded malicious software to another service, say researchers
Center
0 articlesNone
Right
0 articlesNone