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.
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
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.