Contractor’s AI agent on DoD network with CUI

← Back to Blog

Are
there not several issues with this, least of all with a contractor connecting
their personal account so they could use an AI agent to create code using
Claude on a secure network? The DoD system in question contained CUI;
therefore, a contractor’s access to the system is governed not only by
national security laws and regulations, but also by DFARS.

It appears that the Accenture employee’s AI agent ran on the DoD network
for 2 months before it was detected. The agent in question was given admin
access to the DoD environment, could bypass security controls, connected to
the internet, and shared credentials with a 3rd party website. What about
threat detection and admin account monitoring

Pushing any DoD data through AI comes with significant risks. If it was a
personal AI account, then the security of the data cannot be guaranteed, nor
can we know what and how the AI used the data. It’s reported that Accenture
did not report the incident, as they did not believe it was serious, so the
AI agent conformed to DFARS 252.204-7012 for the protection of CUI, had a
system security plan, and was risk-assessed under FISMA as part of
Accenture’s broader system integration.

To quote: “We are in a conflict with Iran right now,” says the
employee, noting that Iran has attempted to hack several federal agencies and
parts of US infrastructure. “It’s the worst time to be having this kind
of cavalier attitude.”

One wonders if this is a FCA breach?

Thaddeus Dziekanowski

Parava Security Solutions

#cybersecurity #nationalsecurity

https://lnkd.in/eAWAqrp2

View Original Post https://www.linkedin.com/feed/update/urn%3Ali%3Ashare%3A7522620695869214721

Categories

Artificial Intelligence, National Security & Defence, Regulation & Compliance

Tags

Verified by MonsterInsights