U.S. agencies to get approved access to Meta’s free AI model Llama

U.S. federal agencies will soon get official access to Meta Platforms’ Llama artificial intelligence system.

The General Services Administration (GSA), which manages federal purchasing, plans to place Llama on its list of approved AI tools. This means that agencies will be able to try the tool under a government safety umbrella, according to procurement lead Josh Gruenbaum.

Gruenbaum said GSA’s decision clears agencies to test Llama, which Meta offers free of charge, with confidence that it satisfies federal security and legal rules. Llama is a large language model that can handle many kinds of data, including text, video, images, and audio.

The green light for Llama follows a broader push by GSA to pre-approve competing systems. In recent months, the agency also validated AI offerings from Amazon Web Services, Microsoft, Google, Anthropic, and OpenAI. As part of those arrangements, the companies agreed to meet federal security requirements and to sell their paid products at sharp discounts, the agency has said.

Asked whether the price cuts were aimed at winning favor with President Donald Trump, Gruenbaum rejected that view. “It’s not about currying favor,” he said. “It’s about that recognition of how do we all lock in arms and make this country the best country it could possibly be.” He said the emphasis is on responsible adoption so that agencies can focus on practical outcomes rather than procurement hurdles.

Expected uses span routine but time-consuming tasks. These include speeding up contract reviews, scanning large documents, and troubleshooting IT issues.

Unlike a traditional contract, this setup did not require a full procurement. Instead, GSA ran a verification process to check that the models meet federal standards. That approach aims to guarantee secure, consistent, and scalable access across agencies, avoiding duplicate efforts and easing the path from pilot to broader use if results prove strong.

Another piece of the arrangement centers on data control. Under the partnership, agencies keep complete ownership of their information while they build tools tailored to their missions. The framework is also meant to lower costs, both through the negotiated pricing and by reusing the same vetted access patterns rather than rebuilding them agency by agency.

Meta remains in the focus of investors

Meta shares are up 33% for the year, topping the Nasdaq 100’s gain of a little over 17%, as investors weigh the company’s heavy spending on AI and the metaverse. Analysts also parsed product reveals from last week’s Connect 2025 event, where Meta unveiled two hardware offerings tied to its broader AI push.

The lineup included the $499 Oakley Meta Vanguard, a device pitched for athletes, and the $799 Ray-Ban Display smart glasses with built-in augmented-reality features. Investor Gene Munster called the Ray-Ban model the “best bang for the buck,” and several analysts voiced optimism even as they cautioned that early adoption can be uneven for new categories.

Bank of America Securities analyst Justin Post said hands-on testing of the Display glasses and the companion Neural Band left him with a positive impression. He described the devices as functional, lightweight, and intuitive. He added that the glasses’ audio-visual performance and on-device AI strengthen his view that AR eyewear could become the key personal device of the AI era.

JPMorgan’s Doug Anmuth argued that Meta’s strength in digital ads gives it room to invest in long-term efforts like AI and the metaverse. He still models Reality Labs operating losses at about $19 billion in 2025 and $21 billion in 2026, and forecasts $71 billion in capital spending next year.

Want your project in front of crypto’s top minds? Feature it in our next industry report, where data meets impact.