Retail incident review.
AI-assisted. Human-decided.
Vatincam is a proposed video analysis service designed to help retailers flag potentially relevant events for human review — with local privacy processing before cloud analysis.
Not a theft determination system. Not a substitute for staff judgment, lawful surveillance practices, or store security procedures.
From footage to a reviewable alert.
Designed around a simple principle: machine-generated signals should prompt a person to check the facts, not trigger automatic accusations or enforcement.
Existing camera footage
Participating stores provide a compatible video source and configure relevant review areas.
Privacy preprocessing
A local model is intended to obscure faces and remove audio before any selected video is uploaded.
AI-assisted analysis
Redacted clips are analysed in the cloud to identify events that may warrant attention.
Staff verification
Authorised staff review the alert and decide whether any action is appropriate.
Privacy designed into the workflow.
These are intended design controls, not a claim that every privacy or security risk has been eliminated. Controls must be validated in testing and deployment.
Face obscuring
Face redaction is intended to run on the store's computer before upload. Redaction can fail on some frames, angles, reflections, or partial faces.
Audio removal
Audio is intended to be stripped locally before cloud processing. The pipeline must verify that no audio track or metadata is transmitted.
Short-lived cloud clips
Selected footage may be stored temporarily in Cloudflare R2 for a configurable short retention period, then automatically deleted. Exact duration is to be published before launch.
Restricted access
Store-scoped permissions, authentication, logging, and secure transfer are planned safeguards, subject to implementation and verification.
Only what's needed
Minimise uploaded footage, fields, and access. Avoid collecting names or identities unless independently justified and lawful.
Clear store notices
Retailers remain responsible for appropriate surveillance notices, camera placement, customer information, and applicable legal obligations.
Simple proposed pricing.
Pricing concept only. GST treatment, included usage, fair-use limits, onboarding, hardware requirements and service availability are not finalised.
Designed for independent retailers
- AI-assisted review alerts
- Local redaction before selected uploads
- Human verification workflow
- Short-lived cloud clip storage
- Store-level account access
Risks, responsibilities & limitations.
No AI surveillance system can guarantee accurate detection, complete anonymity, continuous uptime, or legal compliance in every environment. The following issues must be assessed before a commercial deployment.
False positives
Ordinary shopping behaviour may be flagged incorrectly. Alerts must be checked against context by trained staff; customers must not be accused based solely on a model output.
False negatives
Actual incidents may be missed because of blind spots, lighting, occlusion, camera quality, unusual behaviour or model limitations. Vatincam is not a loss-prevention guarantee.
Imperfect anonymisation
Face blurring does not guarantee anonymity. Clothing, body shape, location, timestamps and other context can still identify a person; faces may be missed.
Data breach or unauthorised access
Cloud storage, credentials, staff devices, integrations or software vulnerabilities could expose clips or metadata. Encryption and access controls reduce but do not eliminate risk.
Privacy & surveillance law
Camera operation, notices, audio recording, employee monitoring, privacy obligations and overseas data handling depend on applicable Australian federal, state and territory laws and the retailer's circumstances.
Wrongful accusations & discrimination
Biased or inaccurate alerts can lead to unfair treatment, reputational damage, disputes and legal claims. The service must not be used for profiling or automatic enforcement.
Service interruptions
Power, internet, camera feeds, local hardware, cloud providers and third-party services can fail. Alerts may be delayed, unavailable or lost.
Retention & deletion
Automatic deletion must be tested, including copies, logs, backups and provider lifecycle rules. Retention schedules and exceptions must be documented; deletion cannot always be instantly guaranteed across every system.
Third-party infrastructure
Cloudflare R2 and any cloud compute or model providers introduce service, security, pricing, regional hosting and contractual dependencies.
Liability, insurance & contracts
Terms should define permitted uses, customer responsibilities, incident response, liability limits, support levels, refunds and insurance. Liability cannot simply be disclaimed in all circumstances.
Local-device compromise
Malware, physical access or misconfiguration on a retailer's computer may expose original unredacted footage before the privacy pipeline runs.
Misuse & overreach
Staff could misuse access or act on alerts improperly. Role-based access, audit logs, training and a clear complaints process are important controls.
Who is responsible for what?
Retailer: lawful camera installation and use; notices and staff/customer communication; camera placement; authorised access; human decisions and on-site responses; compliance with applicable workplace and surveillance rules.
Operator: accurate product representations; reasonable security controls; contractual data handling; access management; incident response; vendor oversight; retention/deletion implementation; responding to applicable privacy and consumer-law obligations. Exact allocation must be defined in signed agreements and cannot override mandatory law.
What happens if footage leaks?
Potential consequences include privacy harm, customer complaints, regulatory inquiries, contractual claims, remediation expenses and reputational damage. The operator should maintain an incident response plan, assess applicable notification duties (including whether the Privacy Act's Notifiable Data Breaches scheme applies), preserve relevant logs, contain access, investigate and notify affected parties where required.
Does local face blurring make the service legally compliant?
No. Blurring is a risk-reduction measure, not a blanket exemption. Other identifiable details can remain, and collecting or using surveillance footage may still be regulated. Independent Australian legal advice and a privacy impact assessment are recommended before launch.
Can staff confront a customer based on an alert?
No automatic conclusion should be drawn. The system should require staff to review source context and follow their employer's lawful policies. Avoid confrontation or adverse action solely because a model generated a flag.
What must be finalised before taking paying customers?
- Document the actual data flow, processors, hosting regions and whether any footage leaves Australia.
- Choose and publish precise retention periods and verify deletion behaviour.
- Test redaction failures, detection accuracy and bias across realistic store conditions.
- Implement security monitoring, least-privilege access, encryption, backups policy and incident response.
- Prepare a privacy policy, customer terms, data-processing agreement where appropriate, and a support process.
- Check Australian privacy, consumer, surveillance and workplace obligations with qualified counsel; review insurance.
- Define GST-inclusive/exclusive pricing, hardware requirements, usage limits and service levels.
This page is a product concept and risk disclosure, not legal advice, a certification, a security audit, or a guarantee of performance. Statements about planned features must be confirmed before commercial release.
Interested in a pilot?
Vatincam is being developed as a responsible, human-in-the-loop retail review service. Commercial availability and technical specifications are subject to validation.
Contact & next steps
Contact details and a public support address will be published before launch. For now, use this page as a product and compliance planning draft.