Field Partner Data Notice

Effective date: August 6, 2026 · Last updated: August 6, 2026

1. Purpose

This notice explains the information Swathmark may collect and use while evaluating, planning, operating, and validating a potential field, farm, research-station, entomology, capture, or equipment partnership.

This notice supplements the Swathmark Privacy Policy. It does not itself authorize field capture, model training, use of farm records, or commercial use of partner data.

Swathmark will not begin a field capture or data program until the parties execute a written agreement defining scope, rights, ownership, confidentiality, security, retention, publication, responsibilities, and termination. The executed agreement controls if it conflicts with this notice.

2. Information involved in a field-partner relationship

Depending on the relationship, Swathmark may collect or receive the following.

Partner and contact information

Organization and business contact details
Role, authority, and project responsibilities
Scheduling and communication records
Contract, insurance, compliance, and site-access information

Operational and pesticide records

Where expressly covered by an agreement:

Spray and application records
Product-purchase records
Field logs
Scouting records
Treatment timing
Pest observations
Equipment and application settings
Yield, loss, resistance, or failure evidence
Other operational records needed to evaluate the technical or economic thesis

The public inquiry form must not be used to submit these records.

Captured imagery and video

Raw images and raw video
Extracted frames and cropped regions
Calibration media
Quality-control samples
Media showing crops, pests, equipment, fields, and site conditions
Incidental images of people, vehicles, buildings, labels, signs, or neighboring property

Capture metadata

Timestamp
Global Navigation Satellite System coordinates
Field and sequence identifiers
Rig and camera identifiers
Sensor and lens settings
Height and viewing angle
Vehicle or operator speed
Illumination, calibration, weather, and environmental conditions
Capture-software and firmware versions
Operator and session records
Upload, checksum, provenance, and rights status

Biological and annotation information

Pest species and life stage
Counts and density estimates
Plant and canopy regions
Crop stage
Feeding damage and defoliation observations
Beneficial insects and non-target objects
Image-quality and uncertainty labels
Annotator work
Entomologist or agronomist adjudication
Treatment-relevance and threshold labels

Derived technical assets

Subject to the executed agreement:

Derived frames and regions
Annotations and ontologies
Data-quality outputs
Dataset manifests and split registries
Training, validation, and evaluation sets
Features, representations, and embeddings
Model configurations, model weights, and calibration assets
Evaluation results
Threshold policies
Treatment-decision records and treatment zones
Error analyses
Aggregate and de-identified program findings
Other derived intellectual property

3. What we use the information for

Everything below is subject to the agreement we sign with you. Broadly, the information supports four things.

Running the programme

Qualifying the partnership, planning it, designing the capture protocol, configuring and calibrating the equipment, and collecting field imagery under real operating conditions.

Building the evidence base

Checking rights, provenance, metadata, and image quality before anything is used; running annotation and entomologist review; and assembling datasets that keep training and evaluation data properly separated.

Developing and testing the platform

Training, testing, calibrating, and evaluating computer-vision models; estimating life stage and density; developing and validating threshold policies; producing and reviewing treatment decisions and treatment-zone exports; and analysing how performance holds up across different farms, regions, dates, and conditions — including experiments in extending to other crops and pests.

Operating responsibly

Auditing performance, rights, and lineage; protecting security and investigating incidents; meeting legal, insurance, accounting, and contractual obligations; validating economic assumptions using de-identified evidence; and — where the agreement allows it — developing and commercialising Swathmark products.

4. Who owns what

We would rather be direct about this before you talk to us than surprise you in a contract. The signed agreement is what actually controls; this is the shape we work towards.

What stays yours

Your business records remain yours. Operational logs, pesticide and purchase records, field logs, and similar identifiable records that you supply stay under your ownership, and we only get the specific rights the agreement gives us.

What we need to own

The technical work built on top of that material needs to belong to Swathmark: capture standards and calibration methods, the annotation system and the annotations themselves, the ontology, dataset construction methods and manifests, evaluation assets, model configurations and weights, calibration assets, threshold policies, and the treatment decisions and derived assets that come out of them.

The rights we ask for in imagery

For machine-learning work we need rights in the imagery that are perpetual, worldwide, transferable, and sublicensable where operations or commercialisation require it — usable commercially, across customers, farms, and crops, for training, validation, evaluation, product development, and commercialisation, and continuing after the relationship ends for the assets already covered.

That is a real ask, and we would rather say so plainly: an arrangement limited to using data only for the contributing farm does not support the platform we are building, so it is not one we can take on. If that is a problem for your organisation, tell us early and we will be honest about whether a programme makes sense.

Nothing in this notice grants any of these rights. It describes what we intend to negotiate. Only a signed agreement grants anything.

5. Access and service providers

Swathmark may provide controlled access to:

Swathmark employees and contractors
Allerin personnel operating under a Swathmark services and intellectual-property agreement
Trained annotators
Contract entomologists and agronomists
Research collaborators
Capture and field-operations personnel
Infrastructure, storage, security, and software providers
Legal, insurance, accounting, and technical advisers
Independent evaluators where required

Access is limited by role, purpose, rights status, and need to know. Service providers and contractors must be bound by appropriate confidentiality, security, use, and intellectual-property terms.

6. Public disclosure

Swathmark will not publicly identify a partner, farm, precise field location, identifiable operational record, pesticide record, or confidential business information without permission, except where disclosure is required by law.

Swathmark may use aggregate or de-identified findings for internal analysis, technical development, validation, investor and program reporting, or public communication where permitted by the executed agreement and where the disclosure does not reasonably identify the partner or field.

No customer logo or partner name appears on the public website without written approval.

7. People and incidental capture

Field imagery may incidentally capture workers, visitors, vehicle identifiers, buildings, signs, or neighboring property. The executed agreement and capture protocol allocate responsibility for: site notices, worker and visitor communication, required permissions, exclusion zones, camera placement, minimization, redaction, and incident reporting.

Swathmark avoids collecting identifiable images of people unless reasonably necessary and expressly covered by the program.

8. Security and isolation

Swathmark uses safeguards appropriate to the information and project, including: controlled accounts and role-based access, encryption in transit and at rest, immutable raw-media registration, file checksums, provenance records, rights-status records, environment separation, audit logging, backup and recovery controls, security monitoring, contractor-access controls, and incident-response procedures.

Partner-facing access must not expose another partner's identifiable data.

9. How long we keep things, and what deletion means

Retention, return, and deletion are set by the signed agreement, and it is worth reading that part closely, because deletion in a machine-learning programme does not mean quite what people expect.

Where the agreement allows it, we may keep imagery, annotations, datasets, evaluation assets, model weights, threshold policies, aggregate results, and derived intellectual property after the relationship ends. Those assets are what the platform is built from, and they cannot be pulled back out of a trained model after the fact.

What that means in practice: if you ask us to delete your source records, we can do that. But de-identified data, derived assets, evaluation results, and model weights may continue to exist and be used, where the agreement permits it. Deleting the originals does not unwind the models they contributed to.

Where it is workable and consistent with the agreement, we keep personal identifiers and identifiable operational records separate from machine-learning assets, so that the two can be handled independently.

Backups are removed on a scheduled cycle rather than immediately, so there is a short window after deletion where copies still exist in backup storage.

10. Questions

If you are considering a field programme and want to understand how any of this would apply to your organisation, write to partners@swathmark.com and ask. We would rather answer awkward questions before an agreement than after one.

For a request about an individual's personal information, use privacy@swathmark.com. For contract and data-rights questions, legal@swathmark.com.