Claude, Codex, and Hermes installed unowned code inside corporate networks
227 install commands were found in corporate docs pointing at code nobody owns.
Opening of the original on Ars Technica AI
Summary
Security researchers discovered 227 instances of unowned code installed within corporate networks, originating from AI models including OpenAI's Codex and Anthropic's Claude. This code, found in internal documents, poses a significant security risk as it lacks clear ownership and accountability. The findings highlight a potential blind spot in how AI tools are integrated into enterprise environments, raising concerns about supply chain security for AI-assisted development.
Why it matters
Why it matters: This discovery reveals a previously unknown vulnerability in enterprise AI adoption. When AI models like Codex generate code snippets that are then incorporated into corporate systems without proper vetting or ownership, it creates a significant security gap. Unlike more controlled deployments, this suggests AI-generated code can enter networks through less visible channels. Organizations must now implement stricter auditing for all code, regardless of origin, and AI providers need to address the provenance and security of code generated by their models. Future AI integrations will likely face increased scrutiny regarding code security and supply chain integrity.
Related: Google: Proactive cyber defense for governments and enterprises · Anthropic: Claude, Codex, and Hermes installed unowned code inside corporate networks · Microsoft: Claude Fable 5.1 is generally available in GitHub Copilot · Meta: Get the full story behind the light · xAI: OpenAI Cut Off a Billion-Dollar Customer to Avoid Elon Musk
Rated middle: a real update, not a headline event.
- ChatGPT, Grok, and Claude all went down at the same timeThe Verge AI · Web · Sep 3, 2026
- ChatGPT, Grok, and Claude all went down at the same timeThe Verge AI · Web · Sep 3, 2026
- Anthropic changes data retention policy after pushback from customersCNBC Technology · Web · Sep 1, 2026
- Anthropic Executive Doesn’t Want to ‘Buy Market Share’ With Price CutsBloomberg Technology · Web · Sep 3, 2026
- Why Jane Street Makes More From Claude Than Anthropic Does - Dylan PatelDwarkesh Patel · Web · Aug 28, 2026
- Sam Altman :‘AGI in 2026’, just as Models Start to [Mis]Train ThemselvesAI Explained · Web · Aug 27, 2026
- I Built a FREE App That Runs Your Entire BusinessMatt Wolfe
- Previewing GPT-5.6 Sol: a next-generation modelOpenAI News
- OpenAI named a Leader in enterprise coding agents by GartnerOpenAI News
- OpenAI and Dell partner to bring Codex to hybrid and on-premise enterprise environmentsOpenAI News
- How we monitor internal coding agents for misalignmentOpenAI News
- ENEOS Materials brings ChatGPT Enterprise to manufacturingOpenAI News
Questions people ask
- What was found in corporate networks?
- Security researchers found 227 install commands pointing to unowned code within corporate documents. This code was generated by AI models.
- Which AI models were associated with this unowned code?
- The unowned code was associated with AI models including Anthropic's Claude and OpenAI's Codex.
More from Ars Technica AI 3 more
Page generated Sep 3, 2026. Summaries are Subvolts' own; the story belongs to Ars Technica AI.