DHS Shifts To Internal AI Restricts Commercial Tools

DHS Mandates Shift from Commercial AI to In House Solutions

Staff at the Department of Homeland Security are now prohibited from using commercial generative artificial intelligence tools such as ChatGPT and Claude. This directive, communicated through a memo to employees this month, marks a significant change in the agency's approach to AI.

The move represents a reversal of a previous policy that had conditionally permitted the use of commercial AI systems. Instead, DHS is pivoting towards technology developed internally.

The Rationale Behind the Policy Reversal

Earlier this month, DHS’s Chief Information Officer, Antoine McCord, issued a memo instructing component tech offices to begin "restricting" the use of generative AI systems. Employees were directed to utilize internal tools instead. Previous guidance on this matter, described by the CIO as "outdated" and "too narrowly" focused on commercial generative AI, has been removed from the agency's online list of IT management directives.

The agency has clarified its stance, with the latest memo explaining that previous approvals for commercial generative AI systems have been revoked. The document defined "Commercial GenAI tools" as "GenAI technology or products available for use or purchase by the general public." It also specified that AI systems created as part of customized software or developed specifically for government use do not fall under this restriction.

Federal AI Adoption Landscape

This decision by DHS unfolds as federal agencies explore various strategies for integrating generative AI into their operations, a recognized priority under both the current and previous administrations. While some government bodies initially blocked access to generative AI systems, Chief Information Officers have gradually begun to formulate usage policies. Agencies like DHS and the General Services Administration have developed their own platforms using commercial technologies. In contrast, others have chosen to employ products such as ChatGPT Gov through government cloud systems.

Introducing DHSChat The Agencys Own AI Tool

A key driver for this policy shift is the agency's focus on its internal tool, DHSChat. Late last year, Michael Boyce, who leads DHS’s AI Corps, announced the availability of DHSChat to nearly 20,000 agency employees. As FedScoop reported at the time, this platform was engineered using commercial large language model technology via an API and built within a secure environment. The introduction of DHSChat followed a series of generative AI pilot programs conducted across DHS component agencies.

Implications for DHS Staff and Future AI Policy

As a consequence of this new direction, DHS staff have been instructed to cancel any existing payments for commercial generative AI systems and to remove related training courses.

Looking ahead, DHS is in the process of developing a new generative AI policy. This development is in accordance with the Trump administration’s recent White House guidance aimed at accelerating the federal government’s adoption and use of AI technology.

This story was supported by the Tarbell Center for AI Journalism.

Madison Alder contributed reporting.