AI Use & Human Oversight Notice
§1Where AI can be used
Depending on the contracted modules and configuration, AI-assisted functionality can include:
- turning a guard’s dictated or written description into a report draft;
- asking focused follow-up questions when a report needs essential context;
- transcribing, translating, normalising or summarising operational text and audio;
- extracting a proposed task, visit or other record from an authorised incoming email or message;
- suggesting labels, priorities, links or next steps for operational review;
- preparing management summaries and identifying records that may need attention; and
- helping authorised users navigate or work with approved SOP content.
Not every feature is enabled for every customer. The Order, deployment configuration and in-product presentation identify the applicable workflow.
§2Transparency at the point of use
Where a person interacts directly with an AI-assisted interface, FieldOps identifies the AI role in the interface or workflow. Drafts and AI-proposed actions are labelled in context where that distinction is material. A website notice alone is not relied upon to explain a live AI interaction.
Some configured workflows perform actions rather than only produce suggestions. Depending on the enabled module, they may add a report comment, close an eligible report, send communications, create a safety incident or trigger escalation. Customer API task writes have separate permissions and confirmation. The configured action scope and resulting history must be reviewed alongside the AI output; this notice does not imply a manual approval before every automated event.
§3Human authority
- Draft before fact: AI-generated text, labels, links and suggested actions must be checked against the underlying situation and available evidence.
- Authorised decision: approval, dispatch, escalation, access, payroll, billing, SOP release and client communication remain governed by the customer’s roles and procedures.
- Correction: authorised users can correct, reject or replace an AI suggestion within the available workflow.
- Traceability: relevant source records, user actions and workflow history can be retained according to the configured module and retention rules.
Customers must train authorised users not to rely on AI output where safety, employment, contractual, financial or legal consequences require independent verification.
§4Workforce and individual impact
FieldOps is not intended to make autonomous decisions about hiring, dismissal, discipline, pay, access denial or other similarly significant outcomes. Operational metrics can be incomplete or reflect context that the system cannot observe.
Any deployment involving individual-level workforce evaluation, profiling, ranking or task allocation based on personal behaviour or characteristics requires a separate written scope and legal review before enablement. That review must address the intended purpose, necessity, lawful basis, worker information, human oversight, retention, contestability and any required DPIA or worker-representation process.
FieldOps is not intended to infer a worker’s emotions. Customers must not repurpose AI output for prohibited or incompatible employment uses.
§5Data used by an enabled AI workflow
Inputs are feature-specific. They may include original text, operational context, conversation history, documents, images or audio. The protected OPS task workflows apply pseudonymisation controls to selected identifiers; this is not a guarantee that all AI inputs across FieldOps are anonymised or pseudonymised. A file, screenshot or recording may contain direct identifiers even when the surrounding request has been minimised. The applicable workflow scope and retention must be documented before use with customer data.
Customer Content is not used to train general-purpose or cross-customer AI models unless the customer explicitly opts in through a separate written agreement. Service telemetry and aggregated reliability information may be used to operate and improve the service as described in the Privacy Notice.
§6AI hosting and international processing
Primary hosting and connected services
The primary FieldOps application and MySQL database are hosted with SiteGround in the Netherlands for the standard EU environment. This statement concerns the primary source environment. It is not a promise that every AI provider, notification service, diagnostic service, backup or support activity processes data exclusively in the EU.
The deployment documentation must identify the relevant storage and access locations, encryption arrangements, key management, backup lifecycle and service providers. A server’s location alone does not establish these facts or remove international-transfer obligations, including those relevant to SenSec LLC as a US legal entity. We do not make a blanket claim that all FieldOps data is encrypted at rest using keys held exclusively in the EU.
Provider retention and legal arrangements
Processing by an external AI service is a separate data flow from storage in the primary FieldOps database. The selected service, API features, retention settings and international-transfer arrangements must be assessed for the actual account and workflow. Conversation history stored in FieldOps and copies retained by a provider have separate lifecycles. We do not describe the standard service as zero-retention or EU-only AI processing.
The signed DPA and completed deployment schedules define the permitted scope. No-training commitments, pseudonymisation and a European contractual entity do not by themselves prove where all processing takes place or replace any safeguard required by data-protection law.
Private AI is a separately assessed deployment
The standard FieldOps service uses external AI services for the workflows agreed for the deployment. A private model, an on-premises environment or a configuration without external AI APIs is not automatically included in that service.
Private AI requirements must be assessed as a separate project. Availability, supported workflows, hosting, network restrictions, support access, fallback behaviour and acceptance tests must be confirmed in writing before a private processing boundary is promised or customer data is used. A request for private AI does not establish that such an environment has already been deployed. Notifications, telephony, messaging, diagnostics and other connected services require their own assessment.
Separate integrations and mobile flows
Customer-selected API and MCP integrations are assessed separately. Authorisation limits the available functions, sites and data, but does not mean that every export passes through the protected OPS pseudonymisation path. The customer must identify an authorised recipient and purpose; the processing roles, retention and applicable transfer arrangements remain relevant.
FieldOps ADMIN notifications are another separate flow. Firebase Cloud Messaging may process notification content outside the EEA. Site names and optional short previews may contain personal or confidential information. A user preference cannot override the organisation’s instructions or an agreed restriction. The app Privacy Policy explains mobile providers, diagnostics, previews and retention. The primary-hosting description above must not be read as an EU-only guarantee for these flows.
See the FieldOps processing notice and private deployment assessment.
§7Known limitations
AI output can be inaccurate, incomplete, stale, biased, overconfident or insensitive to operational context. In particular, an AI system may:
- misread a name, time, place, image, recording or ambiguous instruction;
- produce plausible wording unsupported by the source material;
- miss an exception, contractual requirement or site-specific rule;
- translate specialist language imprecisely; or
- give undue weight to incomplete historical data.
AI output is not legal, HR, medical, safety or regulatory advice. Safety-critical and rights-affecting decisions require an authorised human using appropriate independent evidence.
§8Regulatory classification
AI Act classification is assessed per feature, intended purpose, version and deployment. We do not claim that FieldOps as a whole has one universal risk classification or that a general public page is a Declaration of Conformity.
If a feature is determined to be subject to additional or high-risk requirements in an EU deployment, SenSec may restrict it until the applicable documentation, risk controls, assessment and provider obligations are complete. Customer deployer responsibilities may also apply and are addressed in the deployment documentation.
§9Questions, correction and concerns
An affected person should normally contact the FieldOps customer responsible for the relevant operation or decision. The customer can inspect the source record, review the human decision and ask SenSec for technical assistance under the DPA.
- AI output, bias or harmful useai@sensec.ai
- Privacy and individual rightsprivacy@sensec.ai
- Security vulnerabilitiessecurity@sensec.ai
§10Changes
We update this Notice when the generally available AI workflows or their data-handling boundaries materially change. The signed Order, DPA and deployment documents remain authoritative for a customer; this public page does not silently expand the contracted AI scope.
Related documents
- Data handlingPrivacy Notice · DPA
- EU readinessEU Data Protection & Regulatory Readiness
- Back to overviewLegal & Trust Center