Autonomous Tractors, Gene-Edited Crops and the Emerging Liability Framework for AI in Agriculture
From the CRISPR Patent War to the Right to Repair: How Agricultural Law Is Absorbing Algorithmic Farming
Introduction
At CES 2025, John Deere demonstrated a new generation of autonomous agricultural machines, including the 9RX tractor for large-scale tillage and the 5ML orchard tractor for air-blast spraying, each designed to operate without a human driver in the cab.1 Deere’s second-generation autonomy kit uses computer vision, AI and camera-based perception; the 5ML orchard tractor adds LiDAR sensors for dense orchard canopies. The machines identify obstacles, adjust implement settings and report their progress to a remote operator through a digital interface. Deere has been deploying autonomous capability incrementally across its equipment line since 2022, when it launched the 8R autonomous tractor with tillage capability. By 2026, Deere was marketing autonomous tillage and orchard-spraying capability and extending automation features across additional equipment classes, but availability depends on product line, geography and deployment configuration.1
Outside mining, agriculture is one of the clearest commercial use cases for autonomous heavy equipment. The operational logic is similar: repetitive tasks across large areas, labour shortages and the economic case for precision application of inputs. But the liability framework is different in kind. Mining operates on private land under industrial safety regulation. Agriculture operates on private land adjacent to public roads, residential property and waterways, manages biological systems that respond unpredictably to inputs and generates chemical-drift and environmental-contamination exposures that mining does not. An autonomous mining truck that leaves its designated circuit strikes a manned excavator on the same site. An autonomous tractor that leaves its designated field strikes a cyclist on an adjacent road.
Simultaneously, the genetic engineering of crops has moved from transgenic modification (inserting foreign DNA) to gene editing (modifying the organism’s own genome using CRISPR-Cas9 and related tools). The regulatory and intellectual property frameworks governing gene-edited crops are diverging across jurisdictions in ways that will shape global agricultural trade for decades. The CRISPR patent dispute between the Broad Institute and the University of California, one of the most commercially significant patent disputes in biotechnology, remains commercially unresolved despite the Broad Institute’s victory before the Patent Trial and Appeal Board.2
These two technologies are converging on the same farms. A precision agriculture operation in 2026 may use AI to determine which gene-edited seed variety to plant in which part of a field, deploy autonomous equipment to plant and tend it and use machine-learning models to optimise irrigation, fertilisation and pest management throughout the growing season. The liability framework for this integrated system spans product liability, environmental law, intellectual property, data privacy and the emerging regulation of autonomous vehicles on and adjacent to public roads.
Gene editing is not itself AI, but its commercial deployment increasingly sits inside AI-enabled farming systems: seed selection, field zoning, input optimisation, traceability and yield prediction. The legal significance lies less in any single technology than in the integrated decision chain across seed, software, machinery and data.
Autonomous Agricultural Equipment and Liability
The deployment of autonomous tractors does not simply shift liability from operator to manufacturer. It changes the factual inquiry. A conventional tractor accident usually begins with the conduct of the person operating the machine and the farm business that deployed it. An autonomous-equipment incident would add questions about product design, software performance, remote supervision, field boundaries, obstacle detection, service history and the adequacy of the farm operator’s deployment controls. When an autonomous tractor causes injury, the physical driver may not exist, but there may still be a farm operator, a remote supervisor, a dealer or service network and one or more technology suppliers. The liability analysis does not transfer from one party to another. It layers across the manufacturer (for product defect), the software provider (for navigation or obstacle-detection failure), the farm operator (for the decision to deploy and the adequacy of perimeter safety measures) and, where the machine was remotely supervised, the person exercising the supervisory function.
John Deere’s autonomous systems currently require a human to set up the machine in the field and monitor its operation remotely. The operator can stop the machine at any time through the remote interface. If the machine encounters an obstacle it cannot classify, it stops and alerts the operator. This supervised-autonomy model preserves a human in the loop, but the human’s role is supervisory rather than operational. Whether remote supervision satisfies the duty of care that would apply to a human operator physically present on the equipment has not been tested.
US state law governing agricultural equipment liability varies significantly. Several states have enacted “right to farm” statutes that limit nuisance claims, and in some cases related claims, against established agricultural operations. There appears to be no state-level autonomous-agricultural-equipment liability regime comparable to the EU product-safety model; existing state law instead operates through product-liability, negligence, farm-safety, traffic and right-to-farm principles. The Supporting Innovation in Agriculture Act of 2025 (HR 1705, 119th Congress) proposed tax incentives for investment in innovative agricultural technology projects, including equipment and software used in precision or controlled-environment agriculture, but did not create a safety or liability framework for autonomous agricultural equipment.3
The European Union has legislated directly. The Machinery Regulation (EU) 2023/1230, replacing the 2006 Machinery Directive from 20 January 2027, introduces specific requirements for autonomous mobile machinery: a mandatory supervisory function, defined working areas with physical borders or obstacle detection and a risk assessment that accounts for machines with self-evolving behaviour after placement on the market.7 Safety components with fully or partially self-evolving behaviour using machine-learning approaches and ensuring safety functions fall within Annex I Part A of the Machinery Regulation and require notified-body conformity assessment. The EU AI Act compounds the obligation. Annex I lists both the Machinery Regulation and Regulation (EU) No 167/2013 on agricultural and forestry vehicles. Where an AI system is a product, or a safety component of a product, covered by Annex I Union harmonisation legislation and subject to third-party conformity assessment, it falls into the AI Act’s high-risk regime. In agricultural machinery, the key practical question will be whether the AI element performs a safety function or materially affects safe operation.8 European manufacturers of autonomous agricultural equipment face a layered compliance structure that has no current equivalent in US federal law.
The United Kingdom’s Automated Vehicles Act 2024 addresses automated vehicles on roads and in other public places.9 It does not provide a bespoke regime for autonomous agricultural equipment operating solely on private farmland. An autonomous tractor operating in a field is outside its scope. The same machine crossing a public road to reach an adjacent field enters a different regulatory regime. The regulatory difficulty arises at the boundary between private-field operation and use on roads or other public places.
The CRISPR Patent Landscape
The foundational CRISPR-Cas9 gene-editing patents are the subject of one of the most consequential intellectual property disputes in the history of biotechnology. The Broad Institute of MIT and Harvard holds US patents covering the use of CRISPR-Cas9 in eukaryotic cells (including plant and animal cells), based on the work of Feng Zhang. The University of California, Berkeley, Jennifer Doudna and Emmanuelle Charpentier hold patents covering the CRISPR-Cas9 system itself, based on their earlier work demonstrating the system’s gene-editing capability in vitro.2
The US Patent Trial and Appeal Board ruled in 2022 that the Broad Institute had priority over UC Berkeley for the use of CRISPR-Cas9 in eukaryotic cells. The Federal Circuit remanded the case in May 2025, finding the PTAB had applied the wrong standard. The PTAB reaffirmed its original decision in March 2026. The practical consequence is overlapping patent positions rather than a single clean winner: Broad holds priority for eukaryotic-cell applications in the US interference, while UC-related portfolios remain relevant to the underlying CRISPR-Cas9 system and to other jurisdictions. Commercial applications in plants and animals may require engagement with both patent portfolios, depending on jurisdiction, claim scope, product design and licensing route.2
For agricultural biotechnology companies developing gene-edited crop varieties, the patent landscape creates a licensing cost structure that sits beneath every commercial product. Since 2017, Corteva Agriscience and the Broad Institute have jointly offered non-exclusive licences for agricultural applications of CRISPR-Cas9. Other companies may need to negotiate access through that joint licensing framework or through other portfolio-specific licensing arrangements. Some of that licensing cost may be passed through to farmers through seed pricing, creating an economic dynamic comparable to the technology licensing fees embedded in precision agriculture equipment.
Gene-Edited Crops and Regulatory Divergence
The regulatory treatment of gene-edited crops has diverged sharply across jurisdictions. The United States adopted the USDA’s SECURE rule in 2020, which exempted gene-edited plants from the regulatory oversight that applies to transgenic organisms, provided the genetic modification could have been achieved through conventional breeding. In December 2024, a federal district court in the Northern District of California vacated the SECURE rule, finding that the USDA had failed to articulate a reasoned basis for certain provisions under the Administrative Procedure Act.4 The vacatur was prospective; crops already reviewed under the SECURE process need not be re-reviewed. Following the vacatur, APHIS had to reassess gene-edited plants under the reinstated pre-2020 biotechnology framework, while prior responses and active permits issued before 2 December 2024 remained valid. The practical effect is that the US regulatory posture toward gene-edited crops is now less settled than it appeared twelve months ago.
Europe took the opposite position until recently. The Court of Justice of the European Union ruled in 2018 (Case C-528/16) that organisms produced by gene editing fall within the scope of the EU’s GMO Directive and are subject to its full regulatory requirements.4 This placed gene-edited crops under the EU’s full GMO authorisation regime and materially constrained commercial cultivation. The European Parliament voted in February 2024 to support a revised framework for new genomic techniques. In December 2025, Parliament and the Council reached a provisional agreement distinguishing between NGT-1 plants (equivalent to conventionally bred organisms, exempt from most GMO requirements) and NGT-2 plants (with more extensive modifications, subject to full GMO assessment). The Council adopted the agreed text on 21 April 2026 and the European Parliament adopted the rules on 17 June 2026. Regulation (EU) 2026/1388 was published in the Official Journal on 26 June 2026, enters into force on 16 July 2026 and applies from 17 July 2028, with Articles 29, 30 and 31 applying from 16 July 2026.4
The United Kingdom followed a distinct route. The Genetic Technology (Precision Breeding) Act 2023 received Royal Assent in March 2023. The implementing regulations took effect on 13 November 2025, creating a streamlined authorisation pathway for precision-bred organisms in England where the genetic modification could have been achieved through conventional breeding.10 The Act applies to England only. No equivalent precision-breeding regime currently applies in Scotland, Wales or Northern Ireland. In England, authorised precision-bred food and feed are not treated as GMO products for the purposes of the new regime. In Scotland and Wales, such organisms remain within GMO classification, but UK Internal Market Act principles may allow products lawfully marketed in England to be sold there; UKIMA does not apply to further processing after sale. Northern Ireland is different: under the Windsor Framework, precision-bred products authorised in England cannot currently be sold in Northern Ireland, and precision-bred products may not undergo significant processing, including cultivation, there without satisfying the applicable Northern Ireland/EU regime. The regulatory divergence within a single state adds a further layer of fragmentation to the transatlantic divide.
The cumulative effect is that a gene-edited crop variety may be cleared under a streamlined pathway in England, face an uncertain domestic regulatory path in the United States following the SECURE vacatur, remain classified as a GMO in Scotland and Wales subject to UKIMA market-access issues, be restricted in Northern Ireland under the Windsor Framework, and fall under the new NGT framework in the EU. For global agricultural commodity traders, the compliance burden may include segregation, identity preservation, traceability and jurisdiction-specific documentation for gene-edited product. AI-enabled traceability systems may assist with that task, but the obligation will depend on the product, market and applicable regulatory classification.
The Right to Repair and Data Ownership
The right-to-repair movement in agriculture addresses the question of who controls the equipment after the sale. The dispute arose from the allegation that Deere’s software restrictions limited farmers’ and independent repair providers’ ability to diagnose and repair equipment without Deere-authorised tools, software or dealers.5
In January 2023, the American Farm Bureau Federation and John Deere signed a memorandum of understanding in which Deere committed to making diagnostic and repair tools available to farmers and independent repair providers.5 The MOU did not end the dispute. In January 2025, the Federal Trade Commission, together with Illinois and Minnesota, sued Deere, alleging that its repair restrictions forced farmers towards Deere-authorised dealers. In July 2026, the FTC and five states secured a settlement requiring Deere, for 10 years and under public supervision, to provide farmers and independent repair providers with repair resources equivalent to those available to authorised Deere dealers. That public enforcement settlement followed Deere’s separate USD 99 million private class-action settlement earlier in 2026.5 Several US states have enacted or introduced right-to-repair legislation covering agricultural equipment. Colorado’s Consumer Right to Repair Act (2023) was the first state law to cover agricultural equipment explicitly.5
The right-to-repair question intersects with liability in a direct way. If a farmer repairs autonomous navigation software without manufacturer authorisation and the equipment subsequently causes injury, the manufacturer will argue that the unauthorised modification voids its liability. If the manufacturer restricts repair access in a way that forces the farmer to operate defective equipment while waiting for authorised service, the manufacturer’s restriction may itself ground a claim. The EU’s right-to-repair directive, adopted in 2024, is primarily a consumer-goods measure. It does not create a general commercial repair regime for autonomous agricultural machinery, although the wider repair-access debate will influence agricultural-equipment regulation and contracting.
The data question is equally unresolved. Precision agriculture systems generate vast quantities of operational data: soil conditions, moisture levels, yield maps, input application rates, weather responses. This data has commercial value for seed companies, input suppliers, commodity traders and insurers. The language of “ownership” remains unsettled, but the access position is no longer wholly unregulated. In the EU, the Data Act, which applies from 12 September 2025, gives users of connected products rights to access and share certain data generated by those products.11 That does not resolve every proprietary or competitive question around farm data, but it materially changes the legal analysis for EU-connected agricultural machinery. In the United States, the American Farm Bureau Federation’s Privacy and Security Principles for Farm Data provide a voluntary framework, but voluntary frameworks do not survive a commercial dispute over data access.
AI-Enabled Crop Management and Environmental Liability
AI systems that optimise fertiliser and pesticide application promise to reduce chemical inputs by applying them only where and when they are needed. Variable-rate application technology, guided by AI analysis of drone imagery, soil sensors and crop-growth models, can materially reduce herbicide use in appropriate applications compared with uniform broadcast application.6 The environmental case for precision agriculture is strong.
The liability dimension arises in two directions. If an AI crop-management system under-applies pesticide and a pest outbreak destroys a crop, the farmer may look to the system provider, agronomist or platform vendor, depending on the contract, the recommendation made, the user settings and the extent of any human override. If the system over-applies or misapplies a pesticide and the chemical contaminates an adjacent waterway or harms a neighbouring organic farm, the exposure may involve the farmer who applied the chemical, the platform provider whose recommendation was followed and, in some cases, the chemical manufacturer if the product was applied outside its labelled use parameters or the label warnings were inadequate.
Organic certification is particularly sensitive to chemical contamination. An organic farm that loses its certification because AI-guided spraying on an adjacent conventional farm caused drift contamination may pursue damages that include not only the immediate crop loss but the multi-year revenue impact of lost organic status. The legal categories are familiar — nuisance, negligence, statutory environmental obligations and product-related claims — but proof of causation, drift pathway, algorithmic recommendation, label compliance and certification loss may be heavily contested.
Strategic Outlook
Agriculture’s adoption of autonomous and AI-enabled technology is proceeding under regulatory frameworks that range from layered and technology-specific to fragmented and technology-neutral. The European Union’s layered approach, combining the Machinery Regulation’s autonomous-equipment provisions with the AI Act’s high-risk classification, represents the most structured regime. The United Kingdom has legislated for road vehicles but not for agricultural equipment on private land. The United States has no federal framework for autonomous agricultural equipment at all. Gene-edited crops face a patchwork of national and sub-national regulations that range from England’s streamlined precision-breeding pathway to full GMO-equivalent oversight in Scotland, Wales and Northern Ireland, and, until the NGT regulation applies, the EU. Farm data accumulates in proprietary platforms. Ownership remains unsettled, but access is no longer wholly unregulated: in the EU, the Data Act gives users of connected products rights to access and share certain product-generated data, while the United States still relies largely on contract and voluntary principles.
A serious autonomous-equipment incident would not simply transfer liability from farmer to manufacturer. It would test the allocation of responsibility between the farm operator who deployed the machine, the person supervising it remotely, the manufacturer, the software provider, the dealer or service network and any regulatory regime governing the machine’s operation at the point of incident. Beyond that single event, the convergence of gene-editing and AI-enabled crop management is creating integrated agricultural systems in which a single growing-season decision chain involves intellectual property licences, autonomous equipment, algorithmic input recommendations and regulatory compliance across multiple frameworks. Disputes over access, permitted use and competitive exploitation of farm data are likely to arise as its commercial value increases and the number of parties with platform access multiplies.
Farming is one of law’s oldest commercial environments. It is now absorbing the same autonomous and algorithmic technologies that are transforming shipping, mining, aviation and energy. The liability framework that emerges will reflect agriculture’s distinctive features. Farms are open environments adjacent to public spaces, managing biological systems that respond unpredictably to inputs, under regulatory frameworks that vary between nations and within them. The legal infrastructure is now being asked to allocate risks that were previously absorbed informally by farmers, contractors, dealers and seasonal labour.
Notes
1. John Deere CES 2025 autonomous machines announcement: 9RX tractor (large-scale tillage) and 5ML orchard tractor (air-blast spraying) with second-generation autonomy kit; 5ML adds LiDAR for dense orchard canopies. 8R autonomous tractor with tillage launched CES 2022. By 2026, Deere marketing autonomous tillage and orchard-spraying capability; availability varies by product line, geography and deployment configuration. Computer vision, AI and camera-based perception; remote operator monitoring via Operations Center Mobile (PR Newswire, 6 January 2025; Deere.com; Forbes, 16 January 2025).
2. CRISPR-Cas9 patent dispute: Broad Institute of MIT and Harvard (Feng Zhang) vs University of California, Berkeley (Jennifer Doudna and Emmanuelle Charpentier). PTAB ruled February 2022 that Broad Institute had priority for eukaryotic cell applications. Federal Circuit remanded 12 May 2025, finding PTAB applied wrong conception standard (Berkeley News). PTAB reaffirmed 26 March 2026 (IPWatchdog). Both patent families coexist; agricultural applications may require engagement with multiple patent portfolios, depending on jurisdiction, claim scope, product design and licensing route. Corteva Agriscience and the Broad Institute have jointly offered non-exclusive agricultural licences since October 2017 (Corteva; Simplot press release, 2018).
3. Supporting Innovation in Agriculture Act of 2025, HR 1705, 119th Congress, introduced 27 February 2025; defined precision agriculture technology to include software with data-management and AI systems, GPS guidance, sensors and variable-rate technology (Congress.gov). State “right to farm” statutes provide varying degrees of protection for agricultural operations.
4. USDA SECURE rule adopted 2020 (phased implementation through October 2021), exempting gene-edited plants achievable through conventional breeding from transgenic regulatory oversight. US District Court, Northern District of California, vacated SECURE rule 2 December 2024 under APA (prospective vacatur; crops already reviewed not affected); following the vacatur, APHIS’s pathway for gene-edited plants required renewed analysis under the reinstated biotechnology framework at 7 CFR Part 340 (2019). APHIS programme update; Hogan Lovells analysis. CJEU Case C-528/16 (2018), gene-edited organisms fall within EU GMO Directive. European Parliament vote February 2024 supporting revised NGT framework; Parliament/Council provisional agreement December 2025 distinguishing NGT-1 (exempt from most GMO requirements) and NGT-2 (full GMO assessment). European Parliament press room, Consilium press release reporting. Council adopted the new rules on 21 April 2026 (Consilium press release); Parliament gave final approval on 17 June 2026 (European Parliament press room). Regulation (EU) 2026/1388 published in the Official Journal on 26 June 2026; enters into force 16 July 2026; applies from 17 July 2028, with Articles 29, 30 and 31 applying from 16 July 2026 (EUR-Lex).
5. American Farm Bureau Federation / John Deere memorandum of understanding, signed 8 January 2023, on diagnostic and repair tool access (AFBF; Iowa State CALT). FTC, together with Illinois and Minnesota, sued Deere January 2025 over repair restrictions (FTC press release, 14 January 2025). FTC and five states secured settlement with Deere 8 July 2026 requiring 10 years of equivalent repair-resource access for farmers and independent providers (FTC press release, 8 July 2026). Separate private class-action settlement of USD 99 million earlier in 2026, Deere denying wrongdoing (AP News). Colorado Consumer Right to Repair Act (HB23-1011, 2023), first US state law covering agricultural equipment (Dentons). EU right-to-repair directive adopted 2024.
6. Variable-rate application technology materially reducing herbicide use compared with uniform broadcast application. A 2024 robotic spot-spraying field trial reported 35% average herbicide reduction in sugarcane, with up to 65% reduction in lower-weed-pressure strips (Subeesh et al., arXiv:2401.13931, 2024). AI analysis of drone imagery, soil sensors and crop-growth models for precision input management.
7. Regulation (EU) 2023/1230 on machinery products, published 29 June 2023, replacing Machinery Directive 2006/42/EC from 20 January 2027. New requirements for autonomous mobile machinery include mandatory supervisory function, defined working areas (physical borders or obstacle detection) and risk assessment for machines with self-evolving behaviour after placement on the market. Safety components with fully or partially self-evolving behaviour using machine-learning approaches and ensuring safety functions fall within Annex I Part A and require notified-body conformity assessment (CEMA; EUR-Lex). Harmonised standards list expected by end of 2026.
8. EU AI Act (Regulation (EU) 2024/1689). Annex I lists Regulation (EU) 2023/1230 and Regulation (EU) No 167/2013. Under Article 6, an AI system is high-risk where it is a product, or safety component of a product, covered by Annex I legislation and the relevant product or safety component is subject to third-party conformity assessment. Current Commission guidance states that rules for systems integrated into products such as robotics and industrial machinery apply from 2 August 2028 (European Commission guidelines on high-risk classification; artificialintelligenceact.eu).
9. Automated Vehicles Act 2024 (c. 10), received Royal Assent 20 May 2024. Establishes regulatory framework for automated vehicles on roads and in other public places in Great Britain, covering SAE Level 3 and Level 4 automation. Commencement No. 1 Regulations 2025 (SI 2025/1339). The Act does not provide a bespoke regime for vehicles operating exclusively on private land (legislation.gov.uk; Hogan Lovells analysis).
10. Genetic Technology (Precision Breeding) Act 2023 (c. 6), Royal Assent 23 March 2023. Genetic Technology (Precision Breeding) Regulations 2025, in force 13 November 2025. Creates streamlined authorisation pathway for precision-bred organisms in England where the genetic modification could have been achieved through conventional breeding. Applies to England only; Scotland and Wales retain GMO classification. Precision-bred products lawfully marketed in England may be sold in Scotland and Wales under UKIMA market-access principles, but cannot be grown or substantially processed there without devolved legislation. Northern Ireland remains distinct under the Windsor Framework; precision-bred products cannot currently be sold there under the England-only regime (Defra; Morrison Foerster; Chemistry World; House of Commons Library Research Briefing CBP-9557; UK Explanatory Memorandum on Regulation (EU) 2026/1388).
11. EU Data Act (Regulation (EU) 2023/2854), applicable from 12 September 2025. Gives users of connected products rights to access data generated by those products and to share it with third parties, subject to trade-secret and competitive-sensitivity protections. Applies to agricultural machinery as connected products (European Commission, Digital Strategy).