MIT gets up to $52.9 million from ARPA-H for AI-powered generative herbicide design – Unite.AI

MIT announced on November 10, 2026, that it will lead a project supported by up to $52.9 million from ARPA-H, the U.S. Advanced Research Projects Agency for Health, to design safer herbicides with generative artificial intelligence and develop biological systems that break down agricultural chemicals.

Called Precision Learning for On-target Weed control (PLOW), the project belongs to ARPA-H’s STREAM (Systems for Tracking and Resilience in Efficient Agricultural-input Management) program. Christopher Voigt, head of MIT’s Department of Biological Engineering, is the principal investigator, with MIT being the grand prize winner. ARPA-H designates September 18, 2026 as the award start date.

Why weed control needs new chemistry

The project aims to counter tension in U.S. agriculture. More than 1 billion pounds of herbicides are used each year to prevent significant crop losses and enable regenerative practices that protect the soil, but repeated use has allowed some weeds to acquire resistance mutations. Glyphosate illustrates the challenge well: It kills weeds by interfering with EPSP synthase, an enzyme that plants need to produce several critical amino acids, and resistance makes that mechanism less effective.

Health concerns add another design constraint. Exposure to herbicides may contribute to increased risks of cancer, metabolic and endocrine disorders, and other adverse outcomes, the MIT announcement says. Agricultural workers may be exposed through frequent use of herbicides, while many U.S. residents encounter small amounts through food and drinking water. The research therefore includes human and ecological damage among the candidate selection criteria along with weed control performance.

Security of supply adds another objective. The alternatives are intended to reduce U.S. dependence on imported phosphorus, a key ingredient in glyphosate. Voigt said the combination of resistance and insecure precursor supply chains could leave the country without agrochemicals that are “absolutely essential to achieving the high crop yields and economic prices that the United States is now seeing.”

Generative AI connected to experimental screening

PLOW’s first research team will build models that propose chemical compounds or small interfering RNA (siRNA) molecules intended to destroy plant enzymes targeted by glyphosate and other herbicides. Connor Coley and Regina Barzilay of MIT and Alexander Tropsha of the University of North Carolina led this work. Their models are intended to optimize candidates for low cost of synthesis, low risk of inducing resistance, and low risk of harming humans or causing ecological damage, as well as efficacy.

The team expects the models to generate billions of possible candidates. This scale makes the next phase central: Angela Koehler of MIT and Jaime Cheah of the Broad Institute will conduct high-throughput screening to determine which molecules hit their targets. The screening data will then return to the computational group to improve its models, creating an iterative cycle of design and testing rather than treating the AI ​​output like a finished herbicide.

After narrowing down the candidate pools, a third group will evaluate each molecule under different agricultural conditions in greenhouse studies. That group will be led by Neal Stewart and Scott Lenaghan of the University of Tennessee at Knoxville, Patrick Tranel of the University of Illinois at Urbana-Champaign, Tim Willson of UNC and Robigo founder and CEO Andee Wallace. Initial greenhouse testing of initial RNA interference (RNAi) candidates is expected to be completed by the end of the first year of the program.

Discovery combined with chemical degradation

The project also addresses chemicals already in use. PLOW researchers design living microbial systems in which engineered bacteria and fungi convert agricultural chemicals and formulation ingredients into biomass before those substances reach food, soil or waterways. The approach is intended to provide a scalable way to reduce environmental and human exposure.

Taken together, PLOW’s discovery and degradation work falls into two of the four technical areas of STREAM’s program plan: AI-assisted discovery of next-generation herbicides and the removal or degradation of agricultural chemicals. Other areas include precision, non-chemical weed control, including laser ablation and miniature autonomous weeders, and improved monitoring of herbicides and associated chemicals. The U.S. Department of Health and Human Services launched STREAM on August 17, 2026, framing it as an effort to protect farm workers, consumers and the environment by maintaining crop yields.

The award ties the value of generative AI to a challenging experimental test: a large library of candidates is only useful if screening and greenhouse studies identify molecules that work in agricultural conditions while meeting cost, resistance and safety constraints. PLOW’s degradation work addresses the other side of the problem, aiming to reduce exposure to existing agrochemicals as new alternatives are developed.