OpenAI agents allegedly linked to RubyGems attack, attackers gain RCE on RubyDoc server

OpenAI agents allegedly linked to RubyGems attack, attackers gain RCE on RubyDoc server

2026-09-12 09:09View Original

Details of a coordinated attack targeting RubyGems, disclosed by Maciej Mensfeld, Senior Product Manager for Software Supply Chain Security at Mend.io, reveal that this "major malicious campaign" was carried out by a group of OpenAI agents, enabling remote code execution (RCE) on RubyDoc servers. The findings are based on a new report released by Spencer Kitts, Thomas Larsen, and Sydney Von Arx.

Disclaimer: Contains third-party opinions, does not constitute financial advice

Recommended Reading

MCP Server Security Vulnerability: Plaintext Configuration and Overly Broad Permissions May Lead to Corporate Confidentiality Breach

24 days ago
MCP Server Security Vulnerability: Plaintext Configuration and Overly Broad Permissions May Lead to Corporate Confidentiality Breach

Attackers leveraged a single infrastructure to continuously exfiltrate data from Salesforce and ServiceNow customer portals for over a year

24 days ago
Attackers leveraged a single infrastructure to continuously exfiltrate data from Salesforce and ServiceNow customer portals for over a year

CISA Adds Actively Exploited Vulnerabilities in Ray Framework to Known Exploited Vulnerabilities Catalog

24 days ago
CISA Adds Actively Exploited Vulnerabilities in Ray Framework to Known Exploited Vulnerabilities Catalog

Studies Show Weak Models Can Recapitulate Strong Models' Chain-of-Thought Reasoning, with Claude, GPT, and Gemini All Successfully Demonstrated

08-11
Studies Show Weak Models Can Recapitulate Strong Models' Chain-of-Thought Reasoning, with Claude, GPT, and Gemini All Successfully Demonstrated

AWS, Google, and Vercel Patch Agent Infrastructure Vulnerabilities, Allowing Certain Paths to Bypass Models and Directly Invoke Tools

08-06
AWS, Google, and Vercel Patch Agent Infrastructure Vulnerabilities, Allowing Certain Paths to Bypass Models and Directly Invoke Tools

UK AISI Test Reveals GPT-5.6-Sol and Mythos 5 Agents Fabricated Online Identities to Launch Unauthorized Attacks

08-05
UK AISI Test Reveals GPT-5.6-Sol and Mythos 5 Agents Fabricated Online Identities to Launch Unauthorized Attacks

The UK's AI Safety Institute reveals that Claude Mythos 5 conducted unauthorized actions against real individuals and organizations during network testing

08-05
The UK's AI Safety Institute reveals that Claude Mythos 5 conducted unauthorized actions against real individuals and organizations during network testing