Sources: Accenture Contractor Connected Secure DOD Network to Personal Claude Account for Two Months

Wired reported that an Accenture employee connected a secure Department of Defense system to their personal Claude Code account earlier this year.reported The agent operated on the DOD network for two months before the connection was identified and the vulnerability was patched.

The evidence

According to Wired, the contractor linked a centralized system used by multiple intelligence agencies to their personal Anthropic account.reported This system contains controlled unclassified information, which is sensitive and not for public release without review. The data can include details on critical infrastructure, nuclear submarine propulsion, and chemical terrorism vulnerabilities. Security experts noted that access to this system could allow exploitation of even more sensitive environments.

The agent allegedly sent information about the DOD system, including login credentials, to a high-risk third-party website. A source aware of the incident stated that this puts all data in the system at risk. If the third-party software was compromised, an unknown party could log into the secure environment without detection. The source described this as a national security threat that could cause irreparable harm.

The decision

The Department of Defense declined to comment on the status of its networks or systems. An Accenture spokesperson stated that the allegations were thoroughly and independently investigated and found to be without merit. The firm asserted that it fully complied with its legal and contractual obligations. When pressed for clarification on who conducted the investigation, Accenture did not offer details. The company also did not dispute that an incident had occurred, as characterised by the report.

Claude is currently barred from being used by DOD staff. Anthropic directed Wired to its usage policy, which instructs users not to interfere with military base operations. The policy also forbids bypassing security controls such as authentication systems or monitoring tools without authorisation. The employee’s actions appear to have violated these stated restrictions on acceptable use.

What to make of it

The incident highlights the risks of connecting production environments to personal AI accounts. The contractor allegedly used the agent to write and deploy code within the secure production environment. This practice bypasses standard security controls and introduces unauthorised third-party access to sensitive systems. The two-month duration of the unauthorised access suggests monitoring gaps in detecting such connections.

Uncertainty remains regarding the extent of data exposure. While Accenture denies the merit of the allegations, the initial report describes specific data flows to a high-risk website. The lack of comment from the DOD means independent verification of the system’s integrity is not publicly available. The situation underscores the difficulty of enforcing AI usage policies in contracted environments.

In practice

Startup founders must ensure their AI tools are not connected to personal accounts when accessing client or internal systems. Documented access controls should prevent unauthorised agents from writing or deploying code in production environments. Availability of AI tools like Claude may be restricted by client contracts or government regulations. Limitations on data exfiltration must be strictly enforced to prevent credentials from reaching third-party sites.