Accenture contractor allegedly plugged personal Claude account into US defense system

According to Wired, an Accenture contractor plugged a personal Claude Code account into the US Department of Defense's system and let the AI agent write and deploy code. The agent ran on the defense network for about two months before anyone noticed. During that time, it apparently sent login credentials, passwords, and network details to a third-party website flagged as high risk.

Accenture denies the claims, saying a thorough independent investigation found them baseless and that it met all legal and contract obligations. The company did not say who ran the probe. The Defense Department declined to comment.

The system held controlled unclassified information. Experts warn that such data, when combined, can add up to something more sensitive, and leaked credentials could threaten other systems, though there's no proof that classified systems were breached. Anthropic's policy bans bypassing security controls, and Claude is currently banned for Defense Department staff.