# Pentagon Contractors Face Shadow AI Risks

A Washington Technology opinion article discusses allegations that an Accenture contractor used a personal Claude Code account in a Pentagon production environment; Accenture disputes the allegations. The article argues that slow or restrictive approval pathways can encourage unauthorized “shadow AI” use and recommends making approved AI tools available quickly while applying controls such as agent-specific identities, least-privilege access, reviewed logging, outbound allowlists, and human approval for production deployments. The signal identifies no solicitation, contract award, funding amount, or procurement contact.

- Agencies and contractors can use the recommendations to assess whether AI approval processes and access controls address the risks of production use by AI agents.
- For procurement teams, AI service and system requirements can specify identity, access, logging, outbound-connection, and human-approval controls described in the article.
- Contractors supporting DoD should evaluate how employees obtain and use AI tools in production environments; the reported allegations remain disputed and are not evidence of a confirmed contract action.

**Jurisdictions:** federal
**Industries:** Defense & Military, Information Technology
**Topics:** Artificial Intelligence, Cybersecurity
**Published:** November 09, 2026

### Government Entities
- U.S. Department of Defense (DoD)

### Vendors
- Accenture (Contractor; specific contract role unspecified)

### Sources
- [The Pentagon won't stop rogue AI with another memo - Washington Technology](https://www.washingtontechnology.com/opinion/2026/11/pentagon-wont-stop-rogue-ai-another-memo/417129) - Washington Technology