Trust & Policy Centre

Policies should describe how the system actually operates.

This demonstration contains the working AI+I policy register. Operational policies are being developed from the control architecture. Legal-facing terms and notices require jurisdiction-specific legal review before commercial publication.

Working-draft status

These documents demonstrate the intended governance architecture. They are not representations that every control is already fully implemented, independently audited or legally approved.

Responsible AI & Safety PolicyHuman oversight, bounded autonomy, evaluation, harmful-output and unsafe-action controls.Working operational draftPrivacy & Data Handling PolicyData minimisation, lawful handling, classification, access, cross-border considerations and rights.Legal review requiredRetention & Deletion StandardPurpose-based retention, deletion workflows, meeting/transcript handling and verification.Working operational draftInformation Security PolicyRisk management, access control, encryption, logging, secure engineering and resilience.Working operational draftIncident Response & Breach PlanDetection, containment, escalation, recovery, notification decision logic and lessons learned.Working operational draftBusiness Continuity & Disaster RecoveryBackups, restore tests, RTO/RPO, service continuity and recovery exercises.Working operational draftAcceptable Use PolicyProhibited misuse, unsafe automation, abuse, unauthorised access and customer responsibilities.Legal review requiredRecording & Consent StandardExplicit consent, jurisdiction-aware recording logic, notice, minimisation and meeting retention.Legal review requiredBehavioural & Psychometric Data StandardPurpose limitation, validated instruments, consent/transparency, inference boundaries, contestability and employment safeguards.Working policy · legal/HR review requiredResearch Ethics & Consent PolicyConsent, privacy, data classification, study boundaries, validation and independent review.Working research policySupplier & Subprocessor PolicyProvider due diligence, data handling, model/vendor changes, DPAs and subprocessor register.Working operational draftVulnerability Management & DisclosureReporting channel, triage, remediation, verification and responsible disclosure process.Working operational draftAccessibility & Inclusive Design PolicyAccessible interaction, readable interfaces, keyboard support and inclusive-design testing.Working product policyTerms of Service / Customer TermsLicence, service scope, responsibilities, acceptable use, limitations and commercial terms.Counsel review before useCookie & Tracking NoticeWhat tracking is used, why, preference controls and analytics disclosures.Finalise with production analyticsContent, IP & Provenance PolicyCustomer IP, source provenance, generated-content boundaries, rights metadata and export controls.Working operational draft

Responsible AI & Safety

Human accountability remains central. Agent autonomy should be bounded by explicit permissions, consequence, reversibility and risk. Systems should be evaluated for safety, privacy leakage, factuality, bias or fairness where relevant, refusal quality and task success before consequential deployment.

Consequential actions require proportionate human approval, clear override and escalation paths, change control and post-incident learning.

Privacy & Data Handling

Collect and retain only the data required for the defined purpose. Classify sensitive information, restrict access, document data flows, record subprocessors and apply jurisdiction-specific requirements before production use.

Retention & Deletion

Retention periods should follow purpose, legal requirements and customer commitments. Deletion should be testable, logged and verifiable, including recordings, transcripts, derived content and backups where technically and legally appropriate.

Information Security

Security controls cover identity, MFA, least privilege, secrets, secure SDLC, vulnerability management, logging, supplier risk, backups, incident response and tested recovery. Controls should produce operating evidence rather than exist only as policy text.

Recording & Consent

For ACTIV Seller, live meeting audio, transcripts and behavioural telemetry require explicit consent logic, minimised collection, secure streaming, tenant isolation, retention controls and jurisdiction-aware review before production deployment.

Behavioural & Psychometric Data

Behavioural profiles, assessment results, communication preferences and AI interpretations require a defined purpose, appropriate validation, transparent provenance, access controls, retention boundaries and human review. AI+I separates declared, observed and inferred signals and provides a path for people to correct or contextualise information about themselves.

Research Ethics & Consent

AI+I research should preserve contemporaneous evidence while separating hypotheses from validated findings. Research involving people should define consent, privacy, data classification, access, model/provider records and appropriate ethics or risk review.

Incident Response

AI-specific incident categories include data disclosure, unsafe action, harmful output, model drift, prompt injection, provider compromise and systematic bias. Response should cover detection, containment, escalation, recovery, notification decisions, corrective action and verification that learning changed future behaviour.

Business Continuity & Disaster Recovery

Critical services should have defined recovery objectives, tested backups, restoration procedures, fallback modes and post-exercise learning. Recovery evidence belongs in Intelligence Command.

Acceptable Use

AI+I systems should not be used for unauthorised access, harmful automation, deceptive impersonation, unlawful surveillance, rights infringement or attempts to bypass configured safety and security controls.

Supplier & Subprocessor Governance

Material vendors, models, cloud services and processors should be inventoried and reviewed for security, privacy, retention, training-use terms, residency, continuity and change risk.

Vulnerability Management & Disclosure

Vulnerabilities should be triaged by risk, assigned owners, remediated, retested and closed with evidence. A public reporting channel should be established before broad production exposure.

Accessibility & Inclusive Design

Interfaces should support clear hierarchy, readable contrast, keyboard access, understandable language and reasonable testing across common assistive and mobile contexts.

Content, IP & Provenance

Customer intellectual property, source material, generated outputs and rights metadata should be handled with explicit provenance, permissions, access control and export boundaries appropriate to each product.

Legal documents before launch

Production launch will require jurisdiction-appropriate Terms of Service or customer terms, Privacy Policy, data-processing terms and security schedule where relevant, subprocessor disclosures, recording notices, IP/content rights terms and any sector-specific notices. These should be reviewed by qualified counsel before being treated as final commercial documents.

Cookie & Tracking Notice

The final notice will reflect the actual production analytics and tracking stack. The demonstration does not claim a production tracking configuration that has not yet been selected.