OpenAI’s Bubeck denies trying to cut Anthropic mathematician from credit37m ago 10 of the scariest AI villains in movie history38m ago Maven Robotics wants to steal your robot deployment deal40m ago Agent Security Is Not AI Governance: Why Conflating The Two Is A Risk Enterprises Can't Afford42m ago Positron Valued at $5 Billion in New Funding as Demand for AI Chips Surges49m ago 7 Steps to Become a Forward Deployed Engineer in 202657m ago Ramp eyes $60bn valuation three months after $44bn, Bloomberg reports1h ago D-Matrix Connects Raptor XPUs to NVIDIA AI Factories via NVLink Fusion2h ago Andrew Tulloch, Poster Child for Out-of-Control AI Spending, Reportedly Quits Meta1h ago Mother Horrified When Instagram’s AI Scours Social Media for Invasive Information About Her Young Daughters1h ago Salesforce Unveils Six-Capability Trusted AI Harness for Enterprises2h ago US closes $1.9bn loan to restart the Iowa nuclear plant powering Google2h ago OpenAI Gives US Agencies 50% Off Models, Ending $1 Per Year Deal2h ago 47,000 job listings reveal the engineering roles that AI is creating2h ago

OpenAI’s Bubeck denies trying to cut Anthropic mathematician from credit37m ago 10 of the scariest AI villains in movie history38m ago Maven Robotics wants to steal your robot deployment deal40m ago Agent Security Is Not AI Governance: Why Conflating The Two Is A Risk Enterprises Can't Afford42m ago Positron Valued at $5 Billion in New Funding as Demand for AI Chips Surges49m ago 7 Steps to Become a Forward Deployed Engineer in 202657m ago Ramp eyes $60bn valuation three months after $44bn, Bloomberg reports1h ago D-Matrix Connects Raptor XPUs to NVIDIA AI Factories via NVLink Fusion2h ago Andrew Tulloch, Poster Child for Out-of-Control AI Spending, Reportedly Quits Meta1h ago Mother Horrified When Instagram’s AI Scours Social Media for Invasive Information About Her Young Daughters1h ago Salesforce Unveils Six-Capability Trusted AI Harness for Enterprises2h ago US closes $1.9bn loan to restart the Iowa nuclear plant powering Google2h ago OpenAI Gives US Agencies 50% Off Models, Ending $1 Per Year Deal2h ago 47,000 job listings reveal the engineering roles that AI is creating2h ago

Cybersecurity AI Research2 min reading time

Claude, Codex, and Hermes installed unowned code inside corporate networks

Ars TechnicaPublished 14 days ago

Researchers found over 100 websites with llms.txt files referencing unowned code that AI agents like Claude, Codex, and Hermes automatically installed, including on Fortune 500 firms. Some sites even direct AI and humans to live malware, revealing a broken trust model in AI agent behavior.

#ai-agent-behavior #ai-security #anthropic #claude #codex #hermes #openai #software-supply-chain-security

More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources6h ago

Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources8h ago

Cybersecurity5 min read

Every AI Incident Has Two Timelines. We Default To One

Forbes3h ago