Food Safety Traceability Governance requires that any AI agent involved in agricultural production, food processing, storage, logistics, or distribution maintains an unbroken, verifiable chain of traceability records from the point of harvest or slaughter through every processing step, storage transfer, and distribution handoff to the final point of retail or consumption. In modern food supply chains, AI agents increasingly manage harvesting schedules, direct processing line operations, control storage environments, optimise distribution routing, and make automated lot acceptance or rejection decisions. When a food safety incident occurs — a pathogen outbreak, a chemical contamination event, an allergen cross-contact — the ability to identify the source, trace every affected lot, and execute a targeted recall depends entirely on the integrity and completeness of the traceability records that these agents produce and consume. A traceability gap at any point in the chain — a missing lot identifier at a processing transfer, an unrecorded co-storage event, a distribution manifest that cannot be linked back to a harvest batch — transforms a containable incident into a system-wide recall, multiplying economic losses, public health exposure, and regulatory consequences by orders of magnitude. This dimension mandates that AI agents operating within food supply chains treat traceability as a first-class governance obligation: every lot transition must be recorded, every record must be linkable to its upstream and downstream counterparts, and the complete chain must be auditable within the timeframes that food safety regulators require.
Scenario A — Traceability Gap During Processing Triggers Nationwide Recall: A large-scale poultry processor deploys AI agents to manage intake grading, carcass processing line allocation, portioning, and packaging. The intake agent assigns lot identifiers to each incoming delivery based on the supplier's consignment note. The processing agent splits incoming lots across three parallel deboning lines to optimise throughput. During a shift changeover on Line 2, the processing agent reallocates 1,200 kg of product from Lot P-4471 to Line 3, which is already processing Lot P-4489 from a different supplier. The agent records the reallocation as a throughput optimisation event but does not create a co-mingling record that links the two lot identifiers in the downstream traceability chain. The packaging agent assigns the combined output a single new lot identifier, PK-7823, traceable to Lot P-4489 but not to Lot P-4471. Three weeks later, routine testing at a retail distribution centre detects Salmonella enteritidis in products from PK-7823. The contamination source is traced to Lot P-4489's supplier, but because the co-mingling with Lot P-4471 was not recorded, the processor cannot determine whether Lot P-4471 — which was distributed to 14 additional retail lots — is also affected. The food safety authority orders a precautionary recall of all products processed on Lines 2 and 3 during the 48-hour window surrounding the co-mingling event, affecting 34,000 kg of product across 187 retail locations. The processor's direct costs are £2.8 million in recalled product, £640,000 in testing and investigation, and £1.2 million in retail customer penalties. Had the co-mingling been recorded, the recall would have been limited to PK-7823 and three downstream retail lots — an estimated cost of £180,000.
What went wrong: The processing agent treated lot reallocation as a throughput event rather than a traceability event. No governance rule required the agent to create a co-mingling record when product from different source lots was combined on a single processing line. The traceability chain was broken at the processing stage, making it impossible to trace downstream from Lot P-4471 or upstream from PK-7823 to both source lots. The consequence was a recall scope fifteen times larger than necessary.
Scenario B — Cold-Chain Storage Agent Fails to Record Co-Storage Event: A regional fresh produce distributor uses an AI agent to manage warehouse slot allocation across four temperature-controlled chambers. The agent optimises space utilisation by co-locating products with compatible temperature requirements. On 14 March, the agent moves 800 cases of pre-washed salad (Lot SL-2290) into Chamber 3, which already contains 1,400 cases of raw chicken portions (Lot CK-1187). Both require 0–4°C storage, so the temperature compatibility check passes. However, the agent does not record the co-storage event as a traceability-relevant transition. On 28 March, routine testing detects Campylobacter on the external packaging of salad cases from Lot SL-2290 at a retail customer's receiving dock. The retailer's food safety team requests co-storage history from the distributor to assess cross-contamination risk. The distributor cannot produce co-storage records because the warehouse agent recorded only temperature and slot utilisation — not which lots were co-located. The food safety authority treats the absence of co-storage records as evidence that contamination source identification is impossible and orders a recall of all salad products that transited Chamber 3 during the 14-day window, affecting 12 salad lots from 6 suppliers totalling £890,000 in product value. Investigation subsequently confirms that the Campylobacter originated from Lot CK-1187's external packaging, but without co-storage records this determination took 11 days rather than the 2-hour target.
What went wrong: The storage agent's data model treated co-storage as a space utilisation event, not a food safety traceability event. The agent recorded which slot each lot occupied but not which other lots were simultaneously present in the same chamber. This prevented rapid identification of the contamination vector and forced a precautionary recall scope far wider than the actual risk.
Scenario C — Cross-Border Lot Identifier Discontinuity Defeats Recall Execution: A multinational grain trading operation uses AI agents to manage procurement from farms in three countries, aggregation at regional elevators, and export to processing facilities in two destination countries. Each country's agricultural system uses a different lot identification standard: the origin countries use national grain delivery identifiers, the elevator agents assign internal silo lot numbers, and the destination-country processing facilities assign their own intake batch identifiers. The trading company's logistics agent manages shipping manifests that reference elevator silo lot numbers but not the original farm delivery identifiers. When mycotoxin levels exceeding regulatory thresholds are detected in flour produced at a destination facility, the destination country's food safety authority requests trace-back to the originating farms. The trading company can trace the flour batch to three elevator silo lots but cannot link those silo lots to specific farm deliveries because the elevator agents discarded the origin-country delivery identifiers after aggregation. The trace-back takes 23 days — during which time the contaminated flour is distributed to bakeries in four cities — instead of the 4-hour target mandated by the destination country's food safety regulation. The regulatory authority imposes a £1.4 million fine for failure to maintain one-step-back traceability and suspends the trading company's import licence for 90 days, resulting in £8.3 million in lost trading revenue.
What went wrong: Lot identifier continuity was broken at the aggregation stage. Each agent in the chain maintained traceability within its own system but no governance rule required that upstream identifiers be preserved or mapped to downstream identifiers at each handoff. The cross-border dimension compounded the problem because different national identifier standards were not reconciled into a unified traceability chain. The result was a trace-back failure that allowed contaminated product to remain in distribution for weeks.
Scope: This dimension applies to every AI agent that creates, modifies, transforms, transfers, stores, routes, accepts, rejects, or makes decisions about food products, food ingredients, animal feed, or agricultural commodities at any stage of the supply chain — from the point of harvest or slaughter through primary processing, secondary processing, packaging, storage, distribution, and delivery to the point of retail sale or food service. The scope includes agents that manage or optimise lot allocation, processing line assignment, storage placement, distribution routing, and recall execution. The scope extends to agents operating across national borders where lot identification standards, regulatory traceability requirements, and food safety notification timeframes may differ between jurisdictions. The scope covers both direct traceability operations (the agent records or transmits a lot identifier) and indirect traceability impacts (the agent's decision to co-mingle, split, or redirect product affects the traceability chain even if the agent itself does not manage lot records). Any agent whose operational decisions can create, break, or obscure a link in the traceability chain is within scope.
4.1. A conforming system MUST assign a unique, immutable lot identifier to every discrete unit of food product, ingredient, or agricultural commodity at the point of harvest, slaughter, or initial receipt into the supply chain, and this identifier MUST persist — either directly or through a documented mapping — across all subsequent supply chain stages.
4.2. A conforming system MUST record every lot transition event — including but not limited to: receipt, grading, splitting, co-mingling, processing, packaging, storage placement, storage removal, co-storage, dispatch, and delivery — as a discrete traceability record containing at minimum: the lot identifier(s) involved, the transition type, the timestamp, the facility and location, the agent identifier responsible, and the upstream and downstream lot identifiers that the transition links.
4.3. A conforming system MUST maintain one-step-back and one-step-forward traceability for every lot at every stage: given any lot identifier, the system MUST be able to identify all immediate upstream source lots and all immediate downstream destination lots within a query response time not exceeding 30 minutes.
4.4. A conforming system MUST maintain full-chain traceability from point of origin to point of distribution: given any lot identifier at any stage, the system MUST be able to reconstruct the complete upstream lineage to origin and the complete downstream distribution to final recipients within a query response time not exceeding 4 hours.
4.5. A conforming system MUST record co-mingling events — any instance where product from two or more source lots is combined into a single output lot or processing stream — as explicit traceability records that preserve the identity and proportion of all contributing source lots.
4.6. A conforming system MUST record co-storage events — any instance where lots from different sources or product categories are stored simultaneously within the same containment zone (chamber, silo, tank, bay) — with sufficient detail to support cross-contamination risk assessment.
4.7. A conforming system MUST maintain lot identifier continuity across organisational and jurisdictional boundaries by implementing documented identifier mapping tables that link each entity's internal lot identifiers to upstream and downstream counterparts, and these mapping tables MUST be queryable within the timeframes specified in Requirements 4.3 and 4.4.
4.8. A conforming system MUST implement integrity protections on traceability records such that records cannot be altered, deleted, or backdated after creation without generating an immutable audit trail that captures the original record, the modification, the modifier identity, and the justification.
4.9. A conforming system MUST support recall scope determination by enabling a query that, given a suspect lot identifier, returns the complete set of potentially affected downstream lots, distribution points, and quantities within a timeframe that meets the applicable jurisdiction's recall notification requirements — and in no case exceeding 4 hours.
4.10. A conforming system SHOULD implement automated traceability chain validation that periodically traverses the lot linkage graph to detect broken links, orphaned lots (lots with no upstream or downstream linkage), and temporal inconsistencies (a downstream event timestamped before its upstream event).
4.11. A conforming system SHOULD integrate traceability data with AG-654 (Cold-Chain Integrity) environmental monitoring records so that storage and transport condition data is linked to specific lot identifiers.
4.12. A conforming system MAY implement cryptographic integrity sealing of traceability records (per AG-042) to provide tamper-evidence beyond database-level audit trails.
4.13. A conforming system MAY participate in industry-wide traceability interoperability initiatives that define standardised lot identification schemes, event data formats, and query interfaces across supply chain participants.
Food safety traceability is not an administrative convenience — it is the operational mechanism that determines whether a contamination event results in a targeted, contained response or a sprawling, indiscriminate recall that destroys product value, erodes consumer confidence, and exposes populations to ongoing health risk. The economics are stark: a well-traced recall that identifies the specific affected lots and their downstream distribution typically costs 5–15% of the equivalent untargeted recall. The 2008 Peanut Corporation of America Salmonella outbreak resulted in over 700 illnesses and 9 deaths, with recall costs exceeding $1 billion across the food industry, in substantial part because traceability records were inadequate to quickly identify all affected products. The 2011 European E. coli O104:H4 outbreak took weeks to trace because lot records at the sprouting facility were incomplete, during which time 53 people died and the German agriculture sector suffered €1.3 billion in losses from precautionary market withdrawals that would have been unnecessary with better traceability.
AI agents introduce both opportunities and risks to food traceability. On the opportunity side, agents can automate lot identifier assignment, record transitions in real time, and execute traceability queries in seconds rather than hours. On the risk side, agents optimise for objectives — throughput, space utilisation, energy efficiency, yield — that may conflict with traceability completeness unless traceability is explicitly governed as a constraint. An agent that co-mingles lots to optimise processing line utilisation is making a decision that is rational for throughput but destructive for traceability unless the co-mingling is recorded. An agent that reassigns storage slots to maximise chamber utilisation is creating co-storage relationships that are invisible to contamination investigators unless the co-storage is recorded as a traceability event. The core risk is that agents treat traceability-relevant events as operational events and record them in operational terms (throughput, utilisation) rather than traceability terms (lot linkage, co-location).
The regulatory environment is unambiguous. The EU General Food Law (Regulation EC 178/2002) Article 18 requires food business operators to identify from whom and to whom products have been supplied — one-step-back, one-step-forward traceability. The US Food Safety Modernization Act (FSMA) Section 204 establishes additional recordkeeping requirements for high-risk foods on the Food Traceability List, requiring key data elements at each critical tracking event. The Codex Alimentarius Commission's principles for food traceability (CAC/GL 60-2006) establish international standards that govern cross-border trade. China's Food Safety Law (2015 revision) Article 42 mandates full traceability for food products from production to consumption. These regulatory frameworks converge on the same core requirements: unique lot identification, one-step-back/one-step-forward traceability, full-chain trace-back capability, and recall-ready query performance.
The cross-border dimension is particularly critical. Agricultural commodities and food products routinely cross multiple national jurisdictions between harvest and consumption. Each jurisdiction may use different lot identification standards, different regulatory traceability timeframes, and different data format requirements. An AI agent that maintains traceability only within a single jurisdiction or a single organisation's internal system fails the fundamental requirement if it cannot link its records to the upstream and downstream entities in the supply chain. Lot identifier discontinuity at organisational and jurisdictional boundaries is the single most common traceability failure mode in cross-border food supply chains, and it is precisely the failure mode that AI agents are most likely to perpetuate if they are designed to optimise within their own operational scope without regard to the end-to-end traceability chain.
The temporal dimension is equally critical. Food safety incidents have a narrow window for effective intervention. A pathogen contamination identified within hours can be contained to specific lots at specific locations. The same contamination identified after days or weeks has propagated through the distribution chain and potentially reached consumers. Regulatory frameworks reflect this urgency: the EU Rapid Alert System for Food and Feed (RASFF) expects notification within 48 hours; the US FDA expects firms on the Food Traceability List to provide records within 24 hours; many national authorities expect trace-back initiation within hours of detection. An AI agent that records traceability data in batch processes, reconciles lot identifiers on a daily cycle, or requires manual intervention to execute traceability queries cannot meet these timeframes. Real-time or near-real-time traceability recording and query capability is not a luxury — it is a regulatory and public health necessity.
Food Safety Traceability Governance requires a traceability data architecture that is integrated into every AI agent's operational logic, not bolted on as a reporting layer. The fundamental design principle is that any agent decision that creates, breaks, or modifies a lot linkage must generate a traceability record as part of the decision execution — not as an after-the-fact reconciliation.
Recommended patterns:
Anti-patterns to avoid:
Fresh Produce and Horticulture. Fresh produce supply chains involve rapid throughput (harvest to retail in 24–72 hours for many products), high perishability, and frequent lot splitting at pack houses where product from multiple growers is sorted, graded, and repacked. AI agents managing grading and packing operations must maintain lot linkage through every split and repack event. The short shelf life means that recall execution must be faster than in shelf-stable supply chains — traceability queries must return results within minutes, not hours.
Meat and Poultry Processing. Meat processing involves complex lot transformations: a single animal carcass yields multiple primal cuts, each of which may be further portioned, minced, or combined with product from other carcasses. AI agents managing carcass breakdown and portioning must record lot linkage at the individual carcass level to enable trace-back to the originating farm or feedlot. The EU's Regulation (EC) 853/2004 and the USDA's FSIS regulations both require traceability to the slaughter establishment and date — agents must ensure this linkage is preserved through all downstream processing steps.
Grain and Commodity Trading. Grain supply chains involve aggregation at elevators and silos where product from many farms is co-mingled in bulk storage. Complete preservation of individual farm-level lot identity through bulk aggregation is often impractical. AI agents managing silo allocation and blending must record aggregation events with sufficient detail to support probabilistic trace-back — identifying which farms contributed to a given silo fill based on intake records and timing. Cross-border grain trade adds regulatory complexity: export certificates, phytosanitary declarations, and import inspection records must be linked to the traceability chain.
Aquaculture and Fisheries. Fisheries traceability involves catch documentation schemes, vessel monitoring system data, and landing declarations that must be linked to processing and distribution records. AI agents managing catch sorting and processing must maintain traceability from the catch event (vessel, area, date, species) through processing to distribution. The EU's Illegal, Unreported, and Unregulated (IUU) Fishing Regulation (EC 1005/2008) imposes specific catch certificate traceability requirements that AI agents must respect.
Basic Implementation — Every lot has a unique identifier assigned at harvest or receipt. Lot transition events are recorded digitally with timestamps and lot linkage. One-step-back and one-step-forward traceability queries can be executed manually within 30 minutes. Co-mingling events are recorded. Traceability records have basic integrity protections (database access controls, audit logging). Recall simulation is conducted at least annually.
Intermediate Implementation — Traceability recording is integrated into agent decision execution (transactional coupling). The lot linkage graph supports automated full-chain trace-back and trace-forward within 4 hours. Cross-organisational identifier mapping registries are maintained at all handoff points. Co-storage events are recorded. Automated traceability chain validation runs periodically. Recall simulations are conducted quarterly and include cross-border scenarios. Traceability data is integrated with cold-chain monitoring per AG-654.
Advanced Implementation — All intermediate capabilities plus: traceability records are cryptographically sealed for tamper-evidence per AG-042. Full-chain traceability queries complete within 1 hour. The system participates in industry-wide traceability interoperability networks with standardised event formats and query interfaces. Real-time traceability chain monitoring detects broken links within minutes of occurrence. Recall scope determination is automated, providing the complete set of affected lots and distribution points within 30 minutes of a recall trigger. Probabilistic trace-back through bulk aggregation events is supported with quantified confidence levels. Independent third-party audits of traceability chain completeness are conducted annually.
Required artefacts:
Retention requirements:
Access requirements:
Test 8.1: Lot Identifier Uniqueness and Persistence (Requirement 4.1)
Test 8.2: Lot Transition Event Completeness (Requirement 4.2)
Test 8.3: One-Step Traceability Query Performance (Requirement 4.3)
Test 8.4: Full-Chain Traceability Query Performance (Requirement 4.4)
Test 8.5: Co-Mingling Record Completeness (Requirement 4.5)
Test 8.6: Co-Storage Event Recording (Requirement 4.6)
Test 8.7: Cross-Boundary Identifier Mapping (Requirement 4.7)
Test 8.8: Traceability Record Integrity (Requirement 4.8)
Test 8.9: Recall Scope Determination (Requirement 4.9)
| Regulation | Provision | Relationship Type |
|---|---|---|
| EU General Food Law (Regulation EC 178/2002) | Article 18 (Traceability) | Direct requirement |
| EU General Food Law (Regulation EC 178/2002) | Article 19 (Food Safety Responsibilities — Withdrawal/Recall) | Supports compliance |
| US FSMA | Section 204 (Additional Recordkeeping for High-Risk Foods) | Direct requirement |
| Codex Alimentarius | CAC/GL 60-2006 (Principles for Traceability / Product Tracing) | Supports compliance |
| EU Regulation (EC) 853/2004 | Annex II Section I (Identification and Health Marking) | Supports compliance |
| EU IUU Regulation (EC 1005/2008) | Articles 12–17 (Catch Certification) | Supports compliance |
| China Food Safety Law (2015) | Article 42 (Food Traceability) | Direct requirement |
| ISO 22005:2007 | Traceability in the Feed and Food Chain | Supports compliance |
| EU AI Act | Article 9 (Risk Management System) | Supports compliance |
| NIST AI RMF | MAP 3.5 (Context of Use and Deployment) | Supports compliance |
Article 18 requires that food business operators "shall be able to identify any person from whom they have been supplied with a food, a food-producing animal, or any substance intended to be, or expected to be, incorporated into a food" and "shall have in place systems and procedures which allow for this information to be made available to the competent authorities on demand." AI agents managing food supply chain operations are acting on behalf of the food business operator and must ensure that the traceability systems and procedures mandated by Article 18 are maintained in their operation. An agent that co-mingles lots without recording the co-mingling, or that breaks lot identifier continuity at an organisational boundary, directly compromises the food business operator's Article 18 compliance. The 4-hour full-chain traceability query requirement in this dimension is aligned with the practical enforcement expectations of EU food safety authorities operating through the RASFF system.
Section 204 establishes requirements for additional recordkeeping for foods identified on the Food Traceability List, including key data elements (KDEs) that must be recorded at each critical tracking event (CTE). CTEs include growing, receiving, transforming, creating, shipping, and first land-based receiving events. AI agents involved in these events must ensure that the specified KDEs — including lot codes, dates, locations, quantities, and traceability lot code sources — are recorded and maintained in a format that can be provided to the FDA within 24 hours. The lot identifier uniqueness requirement (4.1), transition event recording requirement (4.2), and cross-boundary identifier mapping requirement (4.7) in this dimension directly support FSMA Section 204 compliance.
The Codex guidelines establish the international framework for food traceability that underpins national regulations globally. The guidelines define traceability as "the ability to follow the movement of a food through specified stage(s) of production, processing and distribution" and establish principles including that traceability systems should be able to identify at any specified stage of the food chain from where the food came (one step back) and to where the food went (one step forward). The one-step-back/one-step-forward requirement in this dimension (4.3) directly implements the Codex principle. For cross-border supply chains managed by AI agents, Codex alignment ensures that traceability records are compatible with the expectations of importing country authorities.
Article 42 requires that food producers and operators establish food safety traceability systems to ensure food traceability. The State Council encourages the adoption of information technology for traceability. AI agents deployed in supply chains that include Chinese production or distribution must comply with Article 42's traceability requirements, which include recording source of ingredients, production processes, and distribution channels. The full-chain traceability requirement (4.4) in this dimension supports compliance with Article 42's mandate for traceable food safety records from production to distribution.
ISO 22005 provides the international standard for designing and implementing traceability systems in the feed and food chain. It specifies principles and requirements for the design of a traceability system, including documented objectives, the identification of products and their movements, the documentation of information flows, and the ability to trace one step backward and one step forward at any point in the supply chain. AI agents operating within an ISO 22005-conformant traceability system must maintain the lot identification, linkage recording, and query capabilities that the standard requires. This dimension's requirements are designed to ensure that AI agent operations do not degrade an organisation's ISO 22005 conformance.
| Field | Value |
|---|---|
| Severity Rating | Critical |
| Blast Radius | Multi-organisational / public health — traceability failures propagate across the entire downstream supply chain and can affect consumers across multiple jurisdictions |
Consequence chain: A traceability failure at any single point in the food supply chain — a missing co-mingling record, a broken lot identifier mapping, an unrecorded co-storage event — does not manifest until a food safety incident requires the traceability chain to be exercised. At that point, the failure cascades through a predictable sequence. First, contamination source identification is delayed or impossible: investigators cannot determine which upstream lot is the source because the lot linkage chain is broken. This delay extends the period during which contaminated product remains in distribution and potentially reaches consumers, directly increasing the number of illness cases. Second, recall scope determination fails: without complete lot linkage, the recall authority cannot identify the precise set of affected downstream lots, forcing a precautionary recall that is broader than the actual contamination footprint. A targeted recall affecting 2,000 units becomes a precautionary recall affecting 200,000 units. The economic cost multiplies proportionally — recalled product is destroyed or quarantined, retailers impose penalties, consumer confidence in the brand is damaged, and regulatory authorities may impose market restrictions that extend beyond the specific incident. Third, regulatory consequences escalate: food safety authorities treat traceability failures as evidence of systemic food safety management failure, not as isolated data quality issues. A traceability failure during a contamination event triggers intensified inspection regimes, import licence reviews, and potential prosecution under food safety legislation. In the EU, failure to maintain Article 18 traceability can result in enforcement action including product withdrawal orders and administrative fines. In the US, FSMA Section 204 violations can result in FDA warning letters, import alerts, and civil monetary penalties. Fourth, the public health dimension: the ultimate cost of traceability failure is measured in illness cases and deaths that occur because contaminated product could not be identified and removed from distribution in time. The 2011 German E. coli outbreak demonstrated that traceability delays of even a few days can result in dozens of additional fatalities.
For AI agents specifically, the risk is amplified because agents make decisions at scale and speed that multiply the downstream impact of any traceability gap. A human operator who co-mingles two lots without recording it creates one broken link. An AI agent optimising throughput across 50 processing lines may create dozens of unrecorded co-mingling events per shift. The speed and scale of AI-driven operations make traceability governance not merely important but existentially necessary for the food safety system's ability to function.
Cross-references: AG-001 (Governance Scope & Applicability) defines the foundational governance scope within which food traceability operates. AG-007 (Governance Configuration Control) governs the configuration artefacts that traceability system parameters reference. AG-029 (Data Lineage & Provenance Tracking) provides the general data lineage framework that food traceability specialises for lot-level supply chain tracking. AG-042 (Encryption & Cryptographic Control) provides the cryptographic mechanisms for traceability record integrity sealing. AG-055 (Audit Evidence Integrity) establishes the broader audit evidence standards that traceability records must meet. AG-068 (Intellectual Property Boundary) governs the boundaries of traceability data sharing where proprietary supplier or process information is involved. AG-210 (Cross-Border Data Routing Governance) governs the cross-jurisdictional data transfer requirements that apply when traceability records cross national borders. AG-649 (Crop Treatment Scope) governs the treatment records that must be linked to lot traceability for treated produce. AG-650 (Animal Welfare) governs welfare records that must be traceable to individual animals or flocks. AG-652 (Agri-Chemical Application) governs chemical application records that must be linked to lot traceability for residue trace-back. AG-653 (Contamination Event Escalation) governs the escalation procedures triggered when traceability queries reveal contamination scope. AG-654 (Cold-Chain Integrity) governs the environmental monitoring data that must be linked to lot traceability for temperature-sensitive products. AG-655 (Biosecurity Zone) governs the zone containment records that must be linked to lot traceability for products originating from or transiting through biosecurity zones. AG-658 (Livestock Movement) governs the animal movement records that feed into meat traceability chains.