Meta calls its DOE lab AI open source. The license says otherwise.

Why Berkeley Lab points a synchrotron at grapevines and lets Meta's models do the reading

The Genesis Mission is a White House-launched Department of Energy initiative from late 2025, built to compress scientific discovery with AI across the national lab system. Its flagship project, SYNAPS-I (SYnergistic Neutron And Photon Science – Intelligence), is led by Lawrence Berkeley National Laboratory with Argonne, Brookhaven, Oak Ridge and SLAC as partners. The problem it's built to solve is a data flood: DOE's light and neutron source facilities now generate tens of petabytes a year, and recent detector upgrades pushed imaging speed from one frame every six seconds to 100,000 frames per second — output no team of human reviewers can keep up with.

At Berkeley Lab's Advanced Light Source, that pipeline is being tested on grapevines. Scientists run micro-CT scans to study drought resilience, and Meta's models — DINOv3 and SAM 3 — now trace the plant's xylem vessels, the tubes that carry water up the stem, without a human drawing a single boundary by hand. Meta describes what comes out the other end as "a fully reconstructed, semantically labeled 3D volume delivered back to the scientist physically standing at the beamline instrument, ready for interpretation." DOE Under Secretary Dario Gil framed the stakes for the whole program: this "compresses discovery time from days to moments and establishes a continuous, self-improving model of science."

The thread's find: a reverse-engineering ban buried in the open-source license

A commenter who pulled the actual SAM 3 and DINOv3 license text went looking for the fine print behind Meta's "open" framing and found conditions a normal open-source license doesn't carry:

"Your use of the SAM Materials will not involve or encourage others to reverse engineer, decompile or discover the underlying components of the SAM Materials."

"You agree not to use, or permit others to use, SAM Materials for any activities subject to the International Traffic in Arms Regulations (ITAR)... including those related to military or warfare purposes, nuclear industries or applications, espionage, or the development or use of guns or illegal weapons."

Their conclusion: "It's always nice when a model's weights are released, but Meta's models are not open source because their weights always come with weird restrictions."

What $280 million actually buys, according to someone who checked

A former B2B/B2G proposal specialist who dug into the Genesis Mission procurement listings put a number on the program the DOE is framing as historic: "Unless someone can correct me, the total amount of grant monies is $280,000,000 or so... That's a pittance and pretty insulting to the purported benefit of funding scientific endeavors." His comparison point was blunt — measured against a single weapons contract, the entire Genesis Mission grant pool barely registers.

The counterargument: a restrictive license, real downstream science

Not everyone in the thread cared about the fine print. A biotech and microscopy researcher pointed to what Meta's earlier open weights already enabled: "Popular microscopy models like Cellpose have leaned heavily on the cornucopia of open and SOTA power... I think it was unthinkable just 5 years ago that a single biologist with just a laptop could do mass-segmentation at this kind of fidelity." Cellpose and the Napari plugin ecosystem, both built on that earlier release, are now standard kit in labs that never had a machine-learning team of their own.

Neither side of the thread actually contradicts the other, and that's the real friction sitting under the Genesis Mission: a consortium of National Labs is now running taxpayer-funded science on a private company's model, under a license written for consumer products, that happens to ban reverse engineering and defense-adjacent reuse. Nobody in the thread, or at the DOE, has had to test yet what happens when a federally funded paper's derivative work bumps into that clause. On a program racing to compress "days to moments," it's the one part of the pipeline nobody has sped up.