AMI: Accessible Private AI Assistant
A friendly, helpful AI assistant built from the ground up for iOS — meeting or exceeding Australian accessibility requirements for digital services used by disabled people — while treating chat privacy as an architectural constraint, not a policy promise. AMI has its own account and subscription; it does not use SOCii Social Services.
Open AMI on ami.socii.auPrivacy architecture
Your conversations are not a product.
Persistent user data stays local. Inference is zero-retention and ephemeral. Chat content is never used for training.
- Persistent data
- Local only
- Inference retention
- Zero
- New-user trial
- 7 days
- Early upgrade
- 25% off
Stored on the user’s device.
Ephemeral processing only.
Full trial before billing.
When upgrading before trial end.
Design principles
Built into the product, not bolted on later.
Accessibility-first interaction
Keyboard, screen-reader, zoom, reflow, target-size, focus visibility and motion preferences are treated as first-order product constraints rather than post-build compliance work.
Local memory, user controlled
Ami can learn useful preferences and context over time without building a remote behavioural dossier. Persistent personal state remains local to the user’s device.
Ephemeral inference
Prompts and responses are used only for the immediate inference transaction under zero-data-retention handling, with no retained server-side chat history by design.
Helpful adaptation without surveillance
The assistant improves through user-visible, local context and tightly minimised anonymous operational analytics — never by secretly monetising or training on private conversations.
Interaction model
A private feedback loop
The product learns where learning is useful, but the persistence boundary stays with the user. Interaction state is visible, reversible and designed not to depend on colour or motion alone.
- 01 · Converse
Ask naturally
Use the assistant through an interface designed for assistive technology, large text, reduced motion, keyboard use and clear interaction states.
- 02 · Adapt
Keep useful context locally
Preferences and user-approved context can improve future sessions while remaining local rather than becoming a remote profile.
- 03 · Infer
Process ephemerally
Only the minimum context required for the current request is sent for inference, under zero-retention handling, then discarded by the inference layer.
Pricing
Clear tiers. Visible capability.
The pricing model is deliberately lower than premium general-purpose AI subscriptions while preserving generous use and the same privacy architecture across both tiers.
Ami
$8/month
Accessible everyday AI with generous usage allowances.
- 7-day free trial for every new user
- Local-only persistent user data
- Zero-retention AI inference
- Generous everyday usage allowances
Upgrade before the end of your trial to receive 25% off your paid subscription.
Ami Pro
$27/month
Higher allowances for users who rely on AI more heavily.
- Everything in Ami
- Substantially higher usage allowances
- Same local-first privacy architecture
- Same zero-retention inference guarantee
The 25% early-upgrade discount also applies to Pro when upgraded before trial end.