Cheaper Weight-Loss Pills Reach U.S. Pharmacies as Insurers and States Strip Coverage

A French startup has compressed multi-year crop breeding cycles into a single season by training a foundation model on 1,600 plant genomes, redirecting agricultural science as drought, heat, and erratic rainfall outrun traditional methods. 

The work signals how artificial intelligence and gene editing have begun reshaping the race to feed a projected 10 billion people by 2050, even as regulators worldwide still disagree on how to classify the crops these tools produce. 

Paris-based Living Models trained its BOTANIC system on more than 1,600 plant genomes, TechRadar reported in late March 2026.

Chief executive Cyril Véran calls BOTANIC a hypothesis engine rather than a decision system, designed to surface biologically coherent genetic combinations instead of statistical correlations.

That framing matters because correlation-only candidates frequently fail in the field, while combinations grounded in actual plant biology survive translation from screen to growing season.

The startup competes against established agrochemical giants such as Bayer CropScience, Corteva, Syngenta, BASF, and Limagrain, which have built internal computational biology teams since the late 2010s.

The transformer-based model studies what scientists describe as the regulatory grammar inside DNA sequences, an approach borrowed from large language models trained on human text.

That structural learning lets the model flag drought-tolerant variants for crops never exposed to current climate stress, a gap traditional breeding has struggled to close.

Traditional drought-resistant variety development can require more than a decade from initial screening to commercial release, a window the company says its system can collapse into months of computation.

BOTANIC outputs calibrated confidence intervals alongside its predictions, flagging low-coverage genomic regions where breeders should treat its recommendations with extra caution.

Identification of useful genes solves only half the problem, because CRISPR-based tools must then edit those targets inside living plants.

Chinese researchers reported 68.6 percent editing efficiency in stable rice lines using a miniaturized Cas12i2Max system, according to CRISPR Medicine News.

In Israel, plant scientists generated 1,300 independent tomato lines through libraries containing 15,804 unique guide RNAs, accelerating disease-resistance screening across the genome.

Seed companies have moved aggressively, with Cibus winning approval in Ecuador for herbicide-tolerant rice and Inari tripling its European greenhouse capacity inside Ghent.

Bayer plans to consolidate its agricultural research operations in Germany by 2028, signaling that even legacy chemistry giants now treat gene-editing pipelines as core infrastructure.

Many of these projects target susceptibility genes rather than spraying chemical pesticides, with the TaMLO gene in wheat offering one of the clearest examples of disease resistance built directly into the seed.

That mechanism creates a fresh regulatory puzzle, because the Global Plant Councilnotes gene-edited plants often display traits indistinguishable from what conventional breeders have produced for decades.

Lawyers and biologists, including Piet van der Meer and Wayne Parrott, have argued that European authorities should treat edits limited to existing plant genes as conventional breeding rather than as genetically modified organisms.

Sweden has carved out specific approval pathways for gene-edited plants, and the United States has signaled it will treat them differently from older GMOs, while the European Union as a bloc still has not finalized that distinction.

Syngenta Vegetable Seeds has partnered with the genome-editing company Tropic to bring CRISPR-derived produce varieties to commercial scale, signaling deeper integration between editing platforms and traditional seed channels.

Living Models target crops first because plant genomes circulate publicly without medical privacy constraints, regulators move faster than the Food and Drug Administration, and field trials deliver feedback inside a single growing season.

Integrated AI-and-CRISPR crop design could improve yields and resource efficiency by 20 to 30 percent, the CARBON newsletter projected.