GSA and Meta launch OneGov initiative to give every federal agency streamlined access to open-source Llama

GSA and Meta streamline government-wide access to Llama open-source AI

The US General Services Administration (GSA) has launched an initiative with Meta to make Llama open-source AI models easier to use across federal departments and agencies. Built under GSA's OneGov strategy, the arrangement eliminates one-off negotiations, speeds up access, and reduces duplicative work across government.

Llama was previously available to US government agencies and contractors supporting national security missions - and even flew aboard the ISS National Laboratory. This initiative extends access government-wide, supporting responsible, cost-effective AI adoption at scale.

Why this matters for your agency

• Faster access to AI: OneGov removes individual procurement hurdles, enabling quicker experimentation and deployment.
• Compliance groundwork done: GSA verified that Llama meets federal requirements and provides consistent, streamlined access.
• Data control by design: Agencies retain full control over data processing and storage, supporting sensitive and mission-critical use cases.
• Lower total cost: Open models reduce licensing overhead and enable teams to build and scale AI applications more affordably.
• Less vendor lock-in: Open availability supports transparency, reproducibility, and flexibility across cloud and on-prem environments.

Policy alignment

This initiative advances objectives in America's AI Action Plan and supports priorities in OMB Memoranda M-25-21 (accelerating federal use of AI with strong governance and public trust) and M-25-22 (driving efficient AI acquisition in government). It helps agencies move from pilots to production with clearer pathways for compliance and oversight.

What's different about this OneGov arrangement

Unlike traditional OneGov agreements, no procurement negotiations were required because Llama is freely available. GSA focused on backend due diligence - validating requirements and setting up consistent access - so agencies can test, adapt, and deploy while maintaining full control over sensitive data.

What leaders are saying

"America is leading on AI and we want to make sure all Americans see the benefit of AI innovation through better, more efficient public services. With Llama, America's government agencies can better serve people," said Mark Zuckerberg, Founder and CEO of Meta.

"Thanks to Meta and its commitment to advancing public sector services, our federal partners now have easier access to Llama and open source AI, which offer unique benefits for federal agencies and government use cases. Through these OneGov initiatives, GSA is driving an unprecedented acceleration of AI adoption across the federal government," added Federal Acquisition Service Commissioner Josh Gruenbaum.

Practical next steps for federal teams

• Define mission use cases where AI can deliver measurable outcomes (e.g., case triage, summarization, pattern detection, knowledge retrieval).
• Establish data protection boundaries early (segmentation, encryption, logging, and audit trails aligned to FISMA/OMB expectations).
• Stand up a sandbox for rapid prototyping with clear exit criteria to production.
• Run structured evaluations (accuracy, bias, safety, cost-per-inference) against baseline workflows.
• Align with agency AI governance: risk assessments, human-in-the-loop policies, and model documentation.
• Plan skills enablement for developers, data owners, security, privacy, and acquisition teams.
• Coordinate with CIO, CISO, Privacy, and Records Management on deployment patterns (on-prem vs. cloud) and retention requirements.
• Track ROI with simple metrics: cycle time reduction, quality gains, cost savings, and user satisfaction.

Bottom line

GSA's OneGov initiative with Meta makes open-source AI more accessible across government, with guardrails that respect agency data control and compliance. Agencies can move faster, spend less, and maintain flexibility as they scale mission-specific AI solutions.

If your team needs structured upskilling on open-source model deployment, AI governance, and evaluation, explore targeted resources at Complete AI Training.