AI improves through training. With scOS, you choose what to contribute.

Why Voluntary AI Training Matters

Artificial intelligence improves through training on real-world data. The more examples AI sees, the better it becomes at recognition, classification, and decision-making. This is fundamental to how modern AI works.

Security camera companies know this. They need training data to improve motion detection, vehicle recognition, scene understanding. The temptation is obvious: millions of customers with cameras capturing useful footage constantly. Free training data at massive scale.

Most companies take this data automatically—justified through vague terms of service language about "improving services" or "product development." Customers don't explicitly consent—they're automatically enrolled through legal text buried in signup processes.

scOS takes a different approach: per-camera, per-model control. When you add a camera, you choose which AI models it contributes to—MotionX (motion detection), ObjecTron (object recognition), or Syne (scene understanding). Each camera has independent settings. Your front drive might contribute to all three models. Your back garden might only help MotionX. Your bedroom cameras might help none.

Your cameras, your choices. AI improvement through granular, transparent contribution—not automatic extraction.

The Automatic AI Training Problem

Most security camera companies train AI on customer footage automatically—with minimal transparency and consent.

Buried in terms of service. Privacy policies include language like "we may analyze customer data to improve our services" or "footage used for product development and AI training." Legal consent through terms you clicked through without reading. No explicit choice about AI training—automatic enrollment justified by legal text.

No transparency about what footage is used. Which cameras contribute footage? All of them—including bedrooms and bathrooms? What events are selected for training? Person detection? Everything that moves? No visibility into what footage is taken from your property for company benefit.

Opt-out is difficult or absent. Want to stop your footage training their AI? Search through settings menus. Maybe there's an opt-out. Maybe it's buried. Maybe it doesn't exist. Some companies make AI training mandatory for using certain features—can't have person detection without contributing to person detection training.

Your data trains models sold commercially. Company develops better AI using customer footage, then licenses those models to other businesses. Your private property becomes unpaid training data for commercial products sold elsewhere. You're providing labor that generates company revenue without compensation or consent.

Changes in training use happen silently. Terms of service update with new AI training purposes. Previously footage was used for person detection training. Now it's training facial recognition, behavioral analysis, demographic classification. New uses applied retroactively to footage already collected—no additional consent required.

How Voluntary AI Training Actually Works

scOS implements ethical AI training with per-camera, per-model control.

Choose when adding each camera. When you add a camera to scOS, you're presented with clear model selection: MotionX (motion detection), ObjecTron (object recognition), and Syne (scene understanding). Choose which models this camera contributes to—or none at all. Each camera has its own settings.

Three specialised AI models. MotionX improves motion detection accuracy—reducing false alerts from shadows, trees, and weather. ObjecTron enhances object and vehicle recognition—better distinguishing cars from bikes from vans. Syne develops scene understanding for context-aware alerts—knowing when activity is unusual for the time and location.

Different cameras, different settings. Your front drive camera might contribute to all three models—it sees vehicles, people, and varied activity. Your back garden camera might only contribute to MotionX—motion detection is useful, but you don't need vehicle recognition there. Your front door camera might contribute to nothing—complete privacy for that location.

Per-camera contribution dashboard. See exactly what each camera contributes: "Front Drive → MotionX: 47 motion events. ObjecTron: 23 vehicle clips (plates masked). Syne: 12 scene clips. Back Garden → MotionX only: 31 motion events." Complete visibility per camera, per model.

Change settings anytime. Added a camera with no contributions but changed your mind? Enable models anytime. Want to stop ObjecTron contributions from your driveway? Disable it for that camera—previously contributed footage deleted from ObjecTron training. Settings always under your control.

Never sold to third parties. MotionX, ObjecTron, and Syne are scOS-exclusive models. Never licensed to other companies. Never sold as training datasets. Your contributions improve scOS for all users—not commercial products sold elsewhere.

Privacy-Preserving Training Techniques

Even when you voluntarily contribute, scOS uses privacy-preserving AI training methods when possible.

Federated learning keeps data local. Instead of centralizing footage for training, federated learning trains models across devices without data leaving properties. Your hub participates in training locally, contributes model updates (not footage) to central server. Learning happens without footage centralization.

Differential privacy prevents individual extraction. Training techniques that ensure individual contributions cannot be extracted from final models. AI learns from aggregate patterns across many contributors—but specific contributors' footage cannot be recovered from trained model. Privacy protection even within training dataset.

Automatic anonymization before contribution. When footage is contributed, privacy-sensitive elements are automatically removed: faces blurred unless facial recognition training specifically opted-in, license plates masked unless ANPR training enabled, audio stripped unless sound detection training selected. Minimize privacy exposure even with voluntary contribution.

Synthetic data reduces real footage needs. scOS invests heavily in synthetic training data—AI-generated footage that looks realistic but contains no real people or properties. Synthetic data reduces how much real customer footage is needed for AI improvement. Less reliance on customer contributions.

Time-limited retention of contributions. Contributed footage used for training, then deleted after model training completes. Not retained indefinitely as permanent training archive. Temporary contribution for specific training purposes, then erased.

What Each Model Does

Each AI model you can contribute to has a specific purpose in making scOS smarter.

MotionX — Motion Detection. MotionX detects movement and distinguishes meaningful motion from noise. Your contributions help it learn: what's a person versus a shadow, wind-blown trees versus actual movement, cats versus concerning activity. Better MotionX means fewer false alerts waking you at 3am for nothing.

ObjecTron — Object Recognition. ObjecTron identifies what's in the frame: people, vehicles, animals, packages. Your contributions help it distinguish: cars from vans from motorcycles, delivery drivers from intruders, your cat from a fox. Better ObjecTron means more accurate "vehicle detected" and "person detected" alerts.

Syne — Scene Understanding. Syne understands context: what's normal for this camera at this time. Your contributions help it learn: delivery activity is normal at 2pm but unusual at 2am, your car leaving at 8am is routine, unfamiliar vehicles lingering is notable. Better Syne means smarter, context-aware alerts.

Combined improvement. The three models work together. MotionX detects movement. ObjecTron identifies what's moving. Syne decides if it's notable given the context. Contributing to all three creates the most value—but contributing to one or two still helps, and contributing to none is completely valid.

Weather and edge cases. All three models benefit from real-world variety: rain, fog, low light, unusual angles, edge cases like wheelchairs or delivery robots. Your specific environment helps train models for conditions similar to yours.

Contributor Benefits and Recognition

scOS appreciates voluntary contributors. Your participation helps make security smarter for everyone—you receive recognition and benefits.

Priority access to beta features. Contributors get early access to new capabilities before general release. Test new AI features first, provide feedback that shapes development. Recognition through prioritized access.

Extended support hours and priority queue. Contributors receive extended customer support hours and prioritized queue. Shorter wait times, more comprehensive assistance. Tangible benefit for your contribution.

Impact visibility. App shows your contribution impact: "Your 87 contributions helped improve person detection accuracy by 0.4% across all users. Your participation made scOS smarter for 12,500 other users." Understand how your contribution helped.

Contributor badge and community recognition. Voluntary contributors receive in-app badge recognizing participation in AI improvement. Optional community recognition if you want public appreciation for helping.

Feature voting influence. Contributors receive weighted votes in feature prioritization surveys. Your input shapes development roadmap more strongly than non-contributors. Influence proportional to participation.

Early problem reporting. If AI makes mistakes that you notice and report, your feedback helps improve models faster. Contributors are essentially beta testers providing real-world validation of AI accuracy.

Transparency About Training Limitations

scOS is honest about voluntary AI training limitations—not all customers will choose to contribute, and that's fine.

Many customers won't participate. Privacy-conscious customers often decline AI training—even with per-camera, per-model control. scOS respects this completely. You're not pressured to contribute. Security works fully with all contributions disabled.

Synthetic data fills gaps. For customers who don't contribute, scOS develops MotionX, ObjecTron, and Syne using synthetic training data and publicly available datasets. Contributors accelerate improvement, but non-contributors aren't disadvantaged—they receive same AI capabilities.

Some models need specific contributions. ObjecTron benefits most from real vehicle clips. Syne needs real-world context variety. MotionX can rely more heavily on synthetic data. Different models have different contribution needs—but all work without any real customer data if necessary.

Regional variations help optimisation. UK weather conditions, British property layouts, and local vehicle types help optimise all three models for British context. But models work globally—UK-specific contributions just make them better for UK conditions.

Each model improves independently. More MotionX contributors means faster motion detection improvement. More ObjecTron contributors means faster object recognition. You can help one model without helping others—every contribution has specific, visible impact.

How Voluntary Training Compares to Alternatives

scOS voluntary AI training differs fundamentally from both automatic training and paid training approaches.

Better than automatic training. Most companies automatically use customer footage. scOS requires explicit opt-in. Better because it respects privacy and gives meaningful choice—not just legal compliance with minimal consent.

Different from paid training. Some companies pay users for training data contributions. scOS provides recognition and benefits instead of payment. Why? Paying creates incentive to contribute footage that might not be appropriate—quantity over quality. Voluntary contribution from users who want to help produces better training data than financially motivated submissions.

More transparent than any alternative. Whether companies train automatically or pay for data, transparency is typically minimal. scOS provides complete visibility into what you contribute, how it's used, what impact it has. Transparency unique regardless of compensation model.

Integration With Other Privacy Features

Per-camera AI training works alongside other scOS privacy capabilities to enable improvement without privacy compromise.

Combined with Encrypted Storage, even voluntary contributions to MotionX, ObjecTron, and Syne use privacy-preserving techniques—footage is anonymised before training use.

Paired with No Data Selling, training data is never sold to third parties—MotionX, ObjecTron, and Syne are scOS-exclusive models, never licensed elsewhere.

Integrated with Transparent Operation, you see exactly what each camera contributes to which models—complete visibility into participation.

Connected to GDPR Compliance, disabling a model for a camera includes deletion of past contributions to that model—satisfying right to erasure even for voluntarily provided data.

The Ethics of AI Training

Artificial intelligence improves through real-world training data. This creates ethical questions: who provides data? Do they consent meaningfully? Do they benefit from improvement their data enabled? Are they exploited or respected?

scOS answers these ethically: per-camera control, per-model selection, transparent dashboards, instant withdrawal, contributor recognition. You choose which cameras help which models. You see exactly what's contributed. You can disable any model for any camera anytime, and past contributions are deleted.

This is ethical AI training: granular contribution from users who choose participation camera by camera, model by model. Not automatic extraction from customers who never explicitly consented.

AI can improve while respecting privacy. Progress can happen through ethical contribution. Technology can advance without exploiting users.

Per-camera, per-model control proves it's possible.

Help Make scOS Smarter—Or Don't

When you add a camera to scOS, you decide: which AI models should this camera contribute to? MotionX for better motion detection? ObjecTron for smarter object recognition? Syne for context-aware understanding? All three? None?

Each camera, each model, your choice. Front drive with all contributions enabled. Back garden with just MotionX. Front door with nothing enabled. Whatever combination makes sense for you.

No automatic enrollment. No all-or-nothing decisions. No pressure to participate. Just per-camera, per-model control with complete transparency about what each camera contributes to which models.

Your cameras, your decision. AI improvement through ethical, granular participation—not automatic extraction.

See all scOS features to understand how Voluntary AI Training works alongside other privacy-focused capabilities to enable improvement without compromising privacy.