Digital Solutions Product Unveil

Nanjing Agricultural University Launches Green Shield, China’s First Open-Source AI Model for Crop Protection

Nanjing Agricultural University (NAU), in partnership with the National Key Laboratory of Agricultural Biosafety and over 30 industry institutions, has launched Green Shield, described as China's first open-source large language model specifically built for crop protection.
Photo by Hongbin on Unsplash

Key Takeaways

  • Nanjing Agricultural University (NAU), in partnership with the National Key Laboratory of Agricultural Biosafety and over 30 industry institutions, has launched Green Shield, described as China's first open-source large language model specifically built for crop protection.
  • The model was trained on a specialized corpus of over 2.5 billion tokens drawn from academic papers, patents, national standards, and field reports covering major crops including rice, wheat, soybeans, vegetables, and fruit trees.
  • Green Shield automatically cross-references China's national pesticide registration database before issuing any recommendation, blocking and self-correcting any non-compliant pesticide advice at the source.
  • The launch follows NAU's February 2026 release of Sinong, China's first open-source vertical LLM for general agriculture, trained on more than 4 billion tokens and available in 8B and 32B parameter versions.
  • NAU has indicated it will continue field testing and model iteration with the goal of producing a tool that is practical and accessible for grassroots farmers.

Nanjing Agricultural University Launches Specialized Crop Protection AI

Nanjing Agricultural University has unveiled Green Shield, an open-source large language model developed specifically for crop protection and pesticide guidance. The model was developed in collaboration with the National Key Laboratory of Agricultural Biosafety and more than 30 industry institutions, and is intended to address a gap in the reliability of AI-generated agricultural advice available to Chinese farmers, according to reporting by Science and Technology Daily and China Daily.

The project was led by Dong Shameng, vice dean of NAU's College of Plant Protection, who described the motivation at the model's launch event.

“China faces frequent crop pest outbreaks and the issue of pesticide resistance. Farmers urgently need professional guidance at the grassroots level. However, general-purpose LLMs often provide inaccurate answers to plant protection questions and, more critically, give poorly standardized, sometimes risky advice on pesticide use,” said Dong Shameng, project leader and vice dean of NAU's College of Plant Protection.

How Green Shield Was Built

To address those shortcomings, the NAU team assembled a domain-specific training corpus of over 2.5 billion tokens, sourced from academic papers, patents, national standards, and field reports. The dataset spans major crop categories including rice, wheat, soybeans, vegetables, and fruit trees, and incorporates information on pest monitoring, green control measures, and pesticide registration data, according to Macau Business.

Wang Dongbo, a professor at NAU's College of Information Management, described the model's core diagnostic and advisory functions.

“With targeted training, the model converges well and recognizes pests with high precision,” said Wang Dongbo, professor at NAU's College of Information Management.

According to Wang, the model is capable of identifying crop types, growth stages, and disease symptoms, and uses that information to generate integrated pest management strategies. Before delivering any recommendation, Green Shield automatically checks each proposed chemical against China's national pesticide registration database, verifying compliance with banned substance lists, approved crop applications, and dosage limits. Non-compliant suggestions are blocked and self-corrected prior to output.

Nanjing Agricultural University's Broader AI Agriculture Push

Green Shield is the second major AI model release from Nanjing Agricultural University in 2026. In February, the university released Sinong at a national higher education forum — described as China's first open-source vertical large language model focused on general agriculture. Sinong was trained on more than 4 billion tokens of specialized agricultural data and is available in 8B and 32B parameter versions via open-source platforms. That release was positioned in support of national strategies on agricultural modernization and artificial intelligence.

Together, the two models represent a concerted effort by NAU to build a specialized AI stack for Chinese agriculture, moving from broad agronomic guidance with Sinong to targeted, compliance-aware crop protection advice with Green Shield.

Next Steps for Green Shield

Wang Yuanchao, vice president of NAU, said the university intends to continue field testing and iterative development of the model.

“Understandable, usable and effective” for farmers, empowering modern agriculture with digital technologies across the entire chain,” said Wang Yuanchao, vice president of NAU, describing the target standard for the model's development.

administrator
As a dedicated journalist and entrepreneur, I helm iGrow News, a pioneering media platform focused on the evolving landscape of Agriculture Technology. With a deep-seated passion for uncovering the latest developments and trends within the agtech sector, my mission is to deliver insightful, unbiased news and analysis. Through iGrow News, I aim to empower industry professionals, enthusiasts, and the broader public with knowledge and understanding of technological advancements that shape modern agriculture. You can follow me on LinkedIn & Twitter.

Leave a Reply

<!-- ============================================================ iGrow Exit-Intent Popup — drop-in snippet Paste this whole block before in your theme footer (footer.php) or via WPCode / "Insert Headers and Footers" plugin → "Site Wide Footer" location. Behavior: - Desktop only (≥768px wide) — skipped on mobile to avoid Google's mobile interstitial penalty. - Triggers (whichever fires first): 1. Exit-intent (mouse leaves viewport top) 2. 15 seconds on page (TIME_MS variable below) 3. 60% scroll depth (SCROLL_PCT variable below) - Frequency cap: once per 30 days per visitor (cookie) - "Remind me later" option: re-shows after 1 day - Excluded paths: /pricing, /checkout, /billing, /account, /subscribe, /login (edit the EXCLUDE regex to change) - Total weight: ~3 KB ============================================================ -->