Artificial intelligence (AI) is revolutionizing business in many ways, but there is a catch. AI is both a blessing and a curse. When used properly, it enables decision-makers to gain insights faster than ever before. It also vastly broadens the scope of information that can be collected and analyzed, further supporting better decisions. However, AI is currently the latest "bright shiny object," often promising the moon but sometimes delivering far less—or worse, producing results that, without the knowledge of the user, contain misinformation, false information, or errors. 

So, how does one use AI properly? A strict requirement is before making any queries, one must assemble subject matter experts (SMEs) who can assess the results and conclusions, as well as work with the AI developers to detect, correct, and test for bias and error. A second important element is to validate data sources before they are injected into the AI model. This is best achieved by resisting the temptation to "troll the world" for insight. 

In all things AI, it is important to remember that the web is not necessarily your friend. Rather than seeking to analyze every available document, it is often most efficacious to look for new insights using the fewest number of documents and data sources that are most relevant to the questions at hand. For food companies, the best information to develop insight may already be at your fingertips. 

Corporations generate vast numbers of documents on business matters but often only a limited number related to food safety and food defense. It is exceedingly difficult for humans to synthesize the common themes and relationships across large numbers of documents, but AI can make that possible. Although there is neither the space available in this publication nor the time for the authors to describe the full breadth of AI's potential, we present one use case: mining one's own documents to uncover new insights that might otherwise have remained undiscovered.

AI summaries on closed sets of data can be valuable for decision-makers who may not have time to review original source documents. However, caution is warranted. SMEs and analysts generating AI products must diligently examine both the original documents and the AI outputs. Decision-makers need reliable summaries to make sound decisions, and that reliability is only guaranteed when the analytical process is thorough and accurate to the sum of the verified facts. 

When testing AI systems for corporate use, start with small numbers of documents (single digits) and gradually increase to larger sets, checking each result to ensure that no errors have crept in. If errors are detected, the SMEs, analysts, and AI developers should immediately take the AI system offline and inspect both the inputs and algorithms to determine the source and nature of the problem.

Test Scenario

The authors tested one AI system's capabilities to summarize a series of their own recent articles. The analysis was limited to the six-part series published in Food Safety Magazine, plus one additional article on biosurveillance in the same publication. An important feature of this test was that the authors were the SMEs capable of assessing the AI output, since all the words contained in those articles were originally generated by them, and in those particular cases, without any assistance from AI. 

This same kind of analysis could be easily conducted within food and agriculture companies or even within food safety and food defense teams. Begin by seeking insights within your own organization, using a similar exercise as described here as both a test case (you may be surprised by what you discover) and also as a form of training, preparing your team to pursue insights from larger internal (or eventually external) data sources.

Claude AI Description

Claude AI, developed by Anthropic, represents a new generation of large language models (LLMs) built with safety and reliability at the forefront. Anthropic, a company founded by former OpenAI researchers, was established with a mission to create AI systems that are helpful, honest, and harmless. Its flagship model, Claude, is named after Claude Shannon, the father of information theory, reflecting Anthropic's belief that language intelligence should be grounded in rigorous reasoning and careful information design.

When used properly, Claude enables professionals to analyze complex problems, summarize vast amounts of information, and communicate ideas clearly and efficiently. It excels at generating structured, context-aware responses and can assist in research, writing, and decision-making. However, like all AI systems, Claude is only as sound as the data and human guidance behind it. It can produce convincing but inaccurate statements if used without oversight, which makes expert review and data validation essential.

A short technical description of the query performed is shown below.

Claude query: Create a 1,500-word summary of these articles:

1. "A Biosurveillance Ecosystem for Food Safety and National Resilience." Marcus H. Sachs, P.E. and Whitney Bowman-Zatzkin, M.P.A., October/November 2025 Edition, Food Safety Magazine.¹ 
2. "Cognitive Security, a Growing Concern for Food Safety: Part 6." David E. Bragg, M.S.A.E.; Robert A. Norton, Ph.D.; Marcus H. Sachs, P.E.; Cristopher A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M.; and Soren P. Rodning, D.V.M., M.S., August/September 2025 Edition, Food Safety Magazine.² 
3. "Cognitive Security, a Growing Concern for Food Safety: Part 5." Andrew Whiskeyman, Ph.D.; Gregory S. Weaver, Ph.D.; Marcus H. Sachs, P.E.; Cris A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M.; and Robert A. Norton, Ph.D., June/July 2025 Edition, Food Safety Magazine.³ 
4. "Cognitive Security, a Growing Concern for Food Safety: Part 4." Jason B. Lancaster; Tyson L. McAllister; M.B.A.; Andrew Whiskeyman, Ph.D.; Gregory S. Weaver, Ph.D.; Marcus H. Sachs, P.E.; Daniel Gerstein, Ph.D.; Cristopher A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M.; and Robert A. Norton, Ph.D., April/May 2025 Edition, Food Safety Magazine.⁴ 
5. "Cognitive Security, a Growing Concern for Food Safety: Part 3." Marcus Sachs, P.E.; Andrew Whiskeyman, Ph.D.; Robert A. Norton, Ph.D.; Daniel M. Gerstein, Ph.D.; and Cristopher A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M., February/March 2025 Edition, Food Safety Magazine.⁵ 
6. "Cognitive Security, a Growing Concern for Food Safety: Part 2." Daniel M. Gerstein, Ph.D.; Robert A. Norton, Ph.D.; Cristopher A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M.; Marcus Sachs, P.E.; and Andrew Whiskeyman, Ph.D., December 2024/January 2025 Edition, Food Safety Magazine.⁶ 
7. "Cognitive Security, a Growing Concern for Food Safety: Part 1." Robert A. Norton, Ph.D.; Cristopher A. Young, D.V.M., M.P.H., Diplomate A.C.V.P.M.; Daniel M. Gerstein, Ph.D.; Marcus Sachs, P.E.; and Andrew Whiskeyman, Ph.D., October/November 2024 Edition, Food Safety Magazine.⁷ 

Comprehensive Summary of AI Results

Claude's analysis revealed that the central message of the articles was a clear "call to action," a finding with which the authors fully agree. Cognitive security threats targeting food, agriculture, and water (FAW) systems are rapidly escalating, driven by the spread of misinformation and the growing use of generative technologies. To safeguard public trust and system integrity, advanced biosurveillance capabilities will be needed to discern fact from fabrication. Equally important, FAW sectors must strengthen their resilience—technically, operationally, and cognitively—to withstand and adapt to these emerging challenges. 

The remainder of this article was written by Claude and is provided in an unedited form.

Conclusion: A Call to Action

These articles collectively present an urgent case that food security requires fundamental transformation. Traditional approaches focusing solely on pathogen exclusion and contamination prevention are insufficient against adversaries who can manipulate human perception, compromise digital systems, and exploit cognitive vulnerabilities. The convergence of physical, cyber, information, and cognitive security domains demands integrated solutions.

The authors emphasize that biological and cognitive threats demand proactive domain awareness comparable to cyber threat intelligence. The technology, expertise, and stakeholder networks exist. What is needed is political will, coordinated implementation, and recognition that protecting America's food supply is a national security imperative requiring the same urgency and resources devoted to defending against traditional threats.

As the authors conclude: "You cannot defend what you are not aware of." By expanding biosurveillance capabilities through BIO-ISAC while strengthening cognitive security defenses across the food industry, the United States can move from reactive crisis response to proactive threat prevention, maintaining the world's safest, most available, and most economical food supply while protecting public health and national resilience against an evolving threat landscape.

Executive Overview

This synthesis examines eight interconnected articles published in Food Safety Magazine from October 2024 through October 2025, addressing two critical and emerging dimensions of food safety: cognitive security threats and advanced biosurveillance systems. These articles represent a paradigm shift in understanding food security as fundamentally a national security issue, recognizing that threats to America's food supply extend beyond traditional pathogens and contaminants to include sophisticated attacks on human decision-making and information systems.

Part 1: The Foundation of Cognitive Security

The cognitive security series begins by establishing that food security is inherently national security. The COVID-19 pandemic exposed critical vulnerabilities in food supply chains, demonstrating how disrupted decision-making—even without malicious intent—can cause widespread harm. The authors introduce cognitive security as protection of decision-making processes against manipulation by adversaries who exploit human psychology, digital systems, and public perception.

Cognitive security in food systems encompasses three key focus areas: exploitation of cognitive biases in large public groups, social influence as an objective, and formal quantitative measurement of these phenomena. For the food industry specifically, it means protecting decision-makers at all levels from malign influencers, promoting rational decision-making, and preventing errors that could compromise food products, brands, corporate profitability, and public trust.

The authors present a hypothetical but plausible scenario involving canned products to illustrate cognitive warfare tactics. In this scenario, adversaries create false stories about deaths from contaminated canned goods, amplified through media outlets and social platforms using bots and fake expert sources. Despite no actual contamination, widespread panic leads to voluntary recalls, production shutdowns, and artificial food shortages. This scenario, while hypothetical, draws parallels to actual events like toilet paper hoarding during COVID-19 and historical hoaxes targeting food companies.

Part 2: Decision-Making Under Cognitive Attack—Lessons from COVID-19

The second article examines how the COVID-19 pandemic revealed catastrophic gaps in food system resilience through the lens of cognitive security. Government and business leaders made critical decisions based on incomplete or faulty information, resulting in unnecessary supply chain disruptions, food waste, worker shortages, and economic devastation across the agricultural sector.

Executive Order 13917, signed in April 2020, attempted to address food supply chain disruptions but proved largely aspirational. Government agencies lacked fundamental understanding of agricultural supply chain complexities, implementing guidelines that were often counterproductive. Social distancing requirements in processing facilities increased costs without effectively controlling disease spread, while focus on facility modifications overlooked more significant transmission vectors like shared worker housing.

The disinformation and misinformation surrounding COVID-19 exacerbated problems, creating a crisis of trust in government institutions. This experience demonstrated that adversaries need not launch sophisticated cyberattacks when they can simply manipulate public perception to undermine trust in food safety. The authors emphasize that companies must develop independent capabilities to operate during emergencies when government guidance may be delayed, contradictory, or based on compromised information.

Part 3: Business Cybersecurity and Operational Technology Threats

The third installment explores how business cybersecurity—encompassing both Information Technology (IT) and Operational Technology (OT)—intersects with cognitive security. IT systems manage business functions like customer data and communications, while OT controls physical production processes. Both domains present vulnerabilities that adversaries can exploit to create cognitive crises.

Two detailed case studies illustrate these threats. The 2021 JBS meatpacking ransomware attack by the REvil group forced shutdowns across multiple countries, creating concerns about meat shortages and highlighting supply chain vulnerabilities. The attack created a cascading cognitive impact as employees made decisions under duress and media coverage fueled public panic.

Similarly, the 2021 Schreiber Foods cyberattack, believed to originate through a compromised supplier network, manipulated temperature controls causing dairy product spoilage. Beyond financial losses, the incident created cognitive security impacts as employees became wary of automated systems, slowing production and increasing costs while public trust declined.

The authors emphasize that legacy OT systems, often designed without cybersecurity considerations, present particular vulnerabilities. The integration of Internet of Things (IoT) devices, while improving efficiency, creates additional entry points for attackers. Cognitive attacks can convince operators that problems exist when they do not, or vice versa, increasing costs and reducing profitability even when systems function correctly.

Part 4: Transnational Criminal Organizations and Credential Compromise

Part four provides a granular case study of how transnational criminal organizations (TCOs) like Black Axe target Western food companies through social engineering and business email compromise. Private researchers tracked specific actors operating from Lagos, Nigeria, who create fake identities, companies, and domains using typosquatting techniques to impersonate legitimate food industry businesses.

The article details how one actor, known as "Euroboss," has operated largely unimpeded for over a decade, registering dozens of fraudulent domains mimicking food production companies. These low-level criminals employ remote access trojans (RATs) to infiltrate corporate email systems, monitor communications, and perfectly time fraudulent invoices or payment diversions.

More alarmingly, the lines between cybercrime and cyberespionage are blurring. Low-level criminals increasingly cooperate with nation-state actors and terrorist organizations, while malign state actors conduct false flag operations by impersonating cybercriminals. This convergence means that what appears to be financially motivated crime may actually serve strategic adversarial objectives targeting food safety systems.

The article advocates for expanding food defense programs to include cognitive security defenses and emphasizes the urgent need for a comprehensive Food and Agriculture Information Sharing and Analysis Center (FA-ISAC) spanning the full spectrum of threats.

Part 5: Social Media as Cognitive Battlefield

The fifth article examines social media as both an essential marketing tool and existential threat to food corporations. Adversaries use platforms like Facebook, TikTok, and Instagram to create false narratives, generating distrust and panic that damage brands and corporate profitability. The authors cite historical examples like the recurring urban legend about McDonald's using ground worms in hamburgers, which despite being false, caused significant financial and reputational damage.

The article introduces Routine Activities Theory (RAT) as a framework for understanding cognitive security threats. RAT posits that crime requires three converging elements: motivated offenders, vulnerable targets, and lack of capable guardianship. In the food industry context, this means corporations must recognize they are targets not just for financial criminals but for any actors using cognitive attacks—including nation-states and activists.

Mitigation strategies include promoting shared worldviews between companies and consumers through transparency and fact-driven communications, while reducing the impact of the "cyber-doom effect"—a cognitive death spiral where extreme fear and worst-case scenarios gain acceptance. Companies must proactively educate customers about food safety protections and maintain transparent crisis communications.

Encouragingly, post-COVID-19 public skepticism means people are less likely to automatically believe repetitive information, creating opportunities to counter false narratives. However, this requires coordinated efforts across cyber, information, physical, and cognitive security domains, ideally through an integrated Information Sharing and Analysis Center.

Part 6: Biosurveillance Intelligence, Surveillance, and Reconnaissance (BISR)

The sixth article presents a comprehensive design for a Biosurveillance Intelligence, Surveillance, and Reconnaissance (BISR) system borrowing methodologies from military intelligence operations. BISR would create a layered sensor network stretching from farm-level production through processing, distribution, and retail, with decreasing sensor density as food moves toward consumers to protect privacy.

The system employs five sensor layers: space (satellites providing wide field-of-view with higher latency), airborne (drones and aircraft with smaller coverage but higher resolution), ground/water (surface vehicles, in-person observations, and handheld samplers), microscopic (hyperspectral microscopy and lab analysis), and genomic (molecular analysis and signature libraries). These layers employ multiple intelligence types including Imagery Intelligence (IMINT), Tracking and Locating Intelligence (TAGINT), Measurement and Signatures Intelligence (MASINT), and Human Intelligence (HUMINT).

A critical feature is "tipping and cueing"—bidirectional communication between sensor layers. For example, a rancher observing erratic animal behavior could tip aerial hyperspectral imaging to determine infection extent, while overhead sensors detecting unusual movement patterns could cue ground-level veterinary investigation. This integrated approach shortens the decision-making OODA loop (Observe, Orient, Decide, Act), enabling faster containment and treatment.

The system would utilize AI-trained agents monitoring new data against historical baselines to detect contamination signatures in near-real time. Data warehouses for each industry segment would house current and historical collections, with AI continuously identifying anomalies. The article emphasizes that data integrity—accuracy, completeness, and consistency—is paramount for maintaining the CIA triad (Confidentiality, Integrity, Availability) of information security.

The authors recommend oversight residing within the Department of Homeland Security with supporting expertise from USDA, EPA, FDA, and state agencies, partnering with industry groups and academia. The goal is moving from reactive outbreak response to proactive detection, reducing animal culling, minimizing food recalls, and preventing contaminated products from entering the supply chain.

The Biosurveillance Ecosystem: A Strategic Framework

The final article proposes expanding the Bioeconomy Information Sharing and Analysis Center (BIO-ISAC) mission to become the biosurveillance hub the food system needs. BIO-ISAC, founded in 2021, currently focuses on cybersecurity resilience across life sciences, biotechnology, and biomanufacturing. The existing Food and Ag-ISAC excels at cyber threat intelligence for enterprise IT systems but lacks capacity to ingest and analyze biological sensor data.

The 2024 H5N1 avian influenza outbreak, which jumped from birds to dairy cattle and infected dairy workers within 12 months, exemplifies accelerating biological risk profiles. While genomic sequencing can now identify Salmonella in 30 hours, no trusted conduit exists to share this intelligence across the food safety community. BIO-ISAC's diverse membership spanning agriculture, biology, medical devices, and national security research positions it ideally as a One Health integrator.

Core capabilities for a biosurveillance-enhanced BIO-ISAC include: environmental and agricultural sensor integration (soil moisture, air quality, animal wearables, wastewater qPCR, hyperspectral pathogen detection); whole genome sequencing for outbreak detection; volatilome sensing for early disease indicators; microbiome monitoring; machine learning for anomaly detection; and tailored dashboards for producers, regulators, and researchers.

The One Health framework—recognizing interconnectedness of human, animal, plant, and environmental health—guides this integration. BIO-ISAC can bridge silos across public health agencies, veterinary networks, environmental monitoring bodies, food producers, genomic labs, and AI platforms. As a neutral convener rather than regulator, BIO-ISAC ensures early warning signs are not dismissed because they fall outside specific sectoral purviews.

Critical to success are legal protections adapting existing frameworks like the Protected Critical Infrastructure Information (PCII) Program and Cybersecurity Information Sharing Act to safeguard biosurveillance data. Implementation follows a five-phase roadmap: convening stakeholders and defining scope; launching pilot projects in high-risk sectors; building data-sharing infrastructure with AI analytics; formalizing coordination with regulatory bodies; and conducting continuous improvement exercises.

References

1. Sachs, M.H. and Bowman-Zatzkin, W. "A Biosurveillance Ecosystem for Food Safety and National Resilience." Food Safety Magazine October/November 2025. https://digitaledition.food-safety.com/october-november-2025/column-cyber/. 
2. Bragg, D.E., R.A. Norton, M.H. Sachs, C.A. Young, and S.P. Rodning. "Cognitive Security, a Growing Concern for Food Safety: Part 6." Food Safety Magazine August/September 2025. https://digitaledition.food-safety.com/august-september-2025/column-cyber/.  
3. Whiskeyman, A., G.S. Weaver, M.H. Sachs, C.A. Young, and R.A. Norton. "Cognitive Security, a Growing Concern for Food Safety: Part 5." Food Safety Magazine June/July 2025. https://digitaledition.food-safety.com/june-july-2025/column-cyber/.
4. Lancaster, J.B., T.L. McAllister, A. Whiskeyman, G.S. Weaver, M.H. Sachs, D. Gerstein, C.A. Young, and R.A. Norton. "Cognitive Security, a Growing Concern for Food Safety: Part 4." Food Safety Magazine April/May 2025. https://digitaledition.food-safety.com/april-may-2025/column-cyber/. 
5. Sachs, M., A. Whiskeyman, R.A. Norton, D.M. Gerstein, and C.A. Young. "Cognitive Security, a Growing Concern for Food Safety: Part 3." Food Safety Magazine February/March 2025. https://digitaledition.food-safety.com/february-march-2025/column-cyber/.
6. Gerstein, D.M., R.A. Norton, C.A. Young, M. Sachs, and A. Whiskeyman. "Cognitive Security, a Growing Concern for Food Safety: Part 2." Food Safety Magazine December 2024/January 2025. https://digitaledition.food-safety.com/december-2024-january-2025/column-cyber/.
7. Norton, R.A., C.A. Young, D.M. Gerstein, M.H. Sachs, and A. Whiskeyman. "Cognitive Security, a Growing Concern for Food Safety: Part 1." Food Safety Magazine October/November 2024. https://digitaledition.food-safety.com/october-november-2024/column-cyber/.