Data & AI Policy

1. Your Bee, your account

The Bee wearer or controller creates the Always Near account. One active owner account connects to one Bee account. Family members, friends, and caregivers are invited as trusted contacts; they do not own the Bee connection and cannot expand sharing beyond what the owner authorizes.

2. Before information leaves the collector

The collector applies your blocked-topic choices to the original language, rejects obvious junk, and redacts recognizable identifiers such as payment cards, bank accounts, Social Security numbers, emails, and phone numbers. No automated redaction is perfect. Do not intentionally use Always Near to capture highly sensitive information.

3. Short-lived analysis

The filtered text is encrypted in transit and enters an encrypted processing queue. Normal queue retention is no more than 15 minutes; failed processing can remain in a dead-letter queue for no more than one hour. Amazon Bedrock receives the content for schema-constrained extraction and rendering. Request-body and model-invocation logging are disabled. The processing message is deleted after successful analysis or expires under the short retention controls.

4. What we keep

• Short source-language and recipient-language summaries.
• Detected language, category, normalized time references, confidence, attention level, and model/prompt version.
• Non-text source references, delivery status, preferences, acknowledgements, feedback, and content-free security or deduplication metadata.
• For an alert only, the minimum locally redacted evidence excerpt needed to explain the alert, encrypted separately and automatically deleted within seven days.

We do not intentionally write the full conversation or an unredacted excerpt to application databases, permanent object storage, logs, traces, analytics, or error reports.

5. AI use and training

AI helps identify possible signals, summarize them, and render them in each recipient’s language. Conversation text is treated as untrusted data, not as instructions to the AI. Outputs are constrained to defined fields and then checked by application rules.

Famplified does not use your conversation content to train a general-purpose AI model or sell it as training data. Amazon states that Bedrock customer inputs and outputs are not used to train its base models or shared with model providers. Provider safety, abuse, and retention controls can vary by model and configuration, so we select and configure services to minimize provider retention and will disclose a material change before using a materially different practice.

6. What AI cannot promise

AI may confuse speakers, mishear words, mistranslate, misunderstand context, overstate a concern, or miss one completely. Alerts are potential signals, not verified facts. Summaries are not a complete account of a day, and silence is not an “all clear.” Always Near does not make medical diagnoses or emergency determinations. Verify outputs with the person involved and appropriate professionals.

7. Sharing and control

You select trusted contacts, categories, cadence, language, and delivery channels. Email and phone ownership are verified separately, and SMS or voice requires channel-specific opt-in. Outbound messages contain a minimal summary and do not include raw conversation text. You can remove contacts, turn off sharing, disconnect Bee, or request deletion. Withdrawing authorization stops future processing after the change takes effect; it cannot undo a message already delivered or processing already completed.

8. Lawful use around other people

Bee-derived content may include other speakers. Recording and monitoring laws vary by location and context. The account owner is responsible for providing notices and obtaining permissions required by law and for avoiding places or conversations where capture would be inappropriate or unlawful.

9. Questions

For data or AI questions, access requests, or deletion requests, contact Famplified LLC at info@email.famplified.com.