Privacy Policy

Last updated: 7 August 2026.

This Privacy Policy explains how Dangerous Animal Alert ("we", "us", "our") collects, uses, and shares information when you use the Dangerous Animal Alert application and related services (the "Service"). Dangerous Animal Alert is operated by Erbacci LTD ("Erbacci"), the controller of personal data described below. Contact: info@erbacciltd.com.

1. What Dangerous Animal Alert does

Dangerous Animal Alert is an awareness tool for wildlife activity near your home. It watches the outdoor Ring cameras you already own (backyard, driveway, or a doorbell facing the yard) and classifies motion events to notify you when a wild animal that could be dangerous to pets — such as a coyote, bear, mountain lion, bobcat, alligator, snake, or bird of prey — is detected, so you can decide whether to bring cats or small dogs inside.

Dangerous Animal Alert detects and informs; it does not protect. It is not a safety device, not a protection system, and not a security or alarm system. It is designed for external, outdoor monitoring only, is never intended for indoor use, and it does not identify people — it performs no facial recognition, reads no license plates, and does no biometric analysis of any kind.

The Service, its interface, and all of its notifications and emails are provided in English only.

2. Information we collect

2.1 Information you provide

• Account identifiers from Ring. When you link your Ring account, we receive your Ring user ID and the email address associated with your Ring account.
• Camera selection and settings. The outdoor Ring camera(s) you choose, the per-species alert toggles you set (which animals you want to be notified about), your notification confidence threshold, and snooze preferences.
• Feedback. Any events you mark "not interesting" or correct ("not a coyote"), used to tune detection thresholds.

2.2 Information we receive from Ring

• Event metadata. Camera motion metadata: camera identifier, timestamp, and a reference to the snapshot frame.
• Snapshot images. A single still frame from your outdoor camera, analyzed to produce a wild-animal species classification only — the species present, a confidence value, and a day/night estimate. The vision models are instructed to classify animals, not people; they do not describe, recognize, or enroll any person. Where and how that analysis happens is set out in §2.2.1 below.

2.2.1 Where and how your snapshots are analyzed

All analysis is performed on infrastructure operated by Erbacci — our cloud environment on Amazon Web Services, and one inference server of our own.

The frame is saved before it is analyzed. When a motion event arrives, we fetch one still frame from Ring and write it to our encrypted storage first, together with a SHA-256 hash of the exact bytes. Only then do the checks below run. This means a frame that turns out to contain no animal at all has still been stored: it is recorded as a suppressed sighting so your timeline shows that motion happened, and it expires on the ordinary timer in §6. We state this plainly because the order matters — nothing here is "checked first and discarded before we keep it".

Then, in order:

1. Animal-presence check — our own inference server, outside the United States. The frame is sent over an encrypted connection (HTTPS/TLS, authenticated with a secret token) to an inference server owned and operated by Erbacci, running an open-weights vision model on our own GPU hardware. That server is reached in two hops: the connection terminates at a reverse proxy running on rented server capacity in Germany, which forwards the frame to the dedicated GPU server itself, which is located in the United Arab Emirates and is where the check is actually performed. The frame therefore transits Germany and is analyzed in the United Arab Emirates — see §10. It is asked a single yes/no question — "is there an animal in this picture?" — and answers with nothing more than a yes/no and a confidence value. It is never asked what species it is, and it is never asked anything about people. If it does not answer clearly, the frame is instead checked by Amazon Rekognition inside our AWS environment. If neither finds an animal, no further analysis happens — but the frame has already been stored, as described above.
2. Species identification — Amazon Bedrock (AWS). Only if an animal is present, the frame is analyzed by Anthropic Claude running on Amazon Bedrock in our AWS environment, in AWS's United States regions, which produces the species classification described above.

What is sent, and what is not. The request sent to our inference server contains the single still frame and nothing else — no name, email address, account identifier, camera name, or any other detail about you — and it is sent only for the binary "is there an animal, yes or no" pre-filter. The species classification never happens there: naming the animal is done exclusively by Amazon Bedrock inside our AWS environment in the United States. It is our own equipment, not an outside company's service: the image is not sent to any third-party AI provider, and the check produces no copy in any product we do not run ourselves.

Your snapshots are never used to train, fine-tune, or improve any AI model — not the model on our own server, and not the models on Amazon Bedrock. All processing is inference-only: the image is analyzed to answer your alert, and that is all.

Your stored snapshot stays in the United States. The saved copy of every sighting lives in encrypted Amazon S3 storage in AWS's US region (us-east-1) and is never moved elsewhere. The brief animal-presence check in step 1 is the one point at which the image is transmitted outside AWS and outside the United States — see §10.

Retention during that check — what we can state, and what we do not claim. Our backend sends the frame for that single call, requests no storage of any kind, accepts only a short yes/no answer in reply, and never logs the image or keeps a second copy of it outside AWS. The server is Erbacci's own equipment, so the frame is never handed to another company at any point on the path. We have not completed an independent verification of what the server's own software writes to disk while it runs the check, and we therefore do not claim that the image is never written there; we will state the verified answer here once that check is done. The only copy we deliberately keep is the stored sighting snapshot, which lives in AWS us-east-1 and is covered by the retention table in §6.

2.3 Information we generate

• Species classifications — the structured JSON output of the vision model: the animal it reports seeing, whether that animal could be dangerous to pets, a day/night estimate, and a confidence value. The model is instructed to identify animals and not people, and we store no other description of the scene — but the animal name is free text the model writes, so we do not claim it is technically impossible for it to contain something else.
• Sighting records — species, severity, scenario, timestamps, and whether the event was pushed or suppressed.
• Prefilter labels. When the Amazon Rekognition fallback runs (see §2.2.1), we store the generic scene labels it returned alongside the sighting. Those labels can include words that describe a person — for example "person", "adult", "child" or "hand" — because Rekognition returns them for any frame containing a person. We keep them only as a record of why the frame was or was not passed on for species analysis. They are never used to identify anyone, never used to describe or judge a person, and never reach the species model's output.
• Snapshot evidence — the still frame stored for a sighting, together with a SHA-256 hash of the exact image bytes so the snapshot can be shown to be untampered.
• Notification status — a single flag on the sighting record recording that a notification was dispatched for it. We do not keep a separate log of which of your devices each alert reached, or whether each individual push succeeded: those per-device attempts and failures appear only in the server-side diagnostic logs described below, and expire on the timer in §6.
• Push notification device tokens and device details — see §2.4.
• Server-side diagnostic and access logs — error traces, system performance metrics, and web-request access logs generated by our own backend. They contain no image or media content. They do contain internal identifiers (your account ID, event and camera IDs) and, in the access logs, the IP address and browser user-agent of requests made to our API. These are developer-side operational and security logs; they are not collected from your device by the app.

2.4 Information from your device

• Push notification tokens (APNs / FCM) so we can deliver alerts.
• Stable per-install device identifier generated by the app on first launch, sent with the push token so we can address the right device. Not linked to any advertising identifier.

The app does not collect analytics or diagnostics from your device, and does not transmit your OS version, app version, or locale for analytics purposes.

We do not collect: faces, faceprints, or other biometric identifiers; license plates; audio; location data outside the camera's fixed outdoor view; contact list, calendar, or other unrelated device data.

3. Bystander privacy, consent, and people who are not users

Dangerous Animal Alert analyzes motion at outdoor cameras you own or control. You are responsible for lawfully operating your Ring cameras and for giving any notice or obtaining any consent that applicable law requires for outdoor recording. Many jurisdictions require that recording devices be disclosed. Dangerous Animal Alert classifies wild animals and does not identify people, but you remain the party responsible for operating the underlying cameras lawfully.

People who are not our users. An outdoor camera can capture people who never signed up for anything — neighbours, delivery drivers, visitors, passers-by, and children. Where a frame contains such a person, we process their image incidentally, as part of answering the camera owner's alert. We want to be clear about what that does and does not involve:

• We make no attempt to identify, recognize, enrol, or describe any person. No facial recognition, no faceprints, no biometric analysis of any kind, and no licence-plate reading.
• The species model's instructions forbid identifying or describing any person, and its output schema has room only for animal fields.
• The image is retained on the ordinary timer in §6 and is visible only to the signed-in account holder whose camera captured it.
• Where EEA or UK law applies, we rely on legitimate interests (Article 6(1)(f) GDPR) for this incidental processing — our interest, and the camera owner's, in being told that a dangerous animal is present. A written assessment of that balance is being prepared.

If you believe you were captured by a camera using Dangerous Animal Alert and you want to exercise your rights, write to info@erbacciltd.com. Please note that we cannot search our systems by a person's appearance — we have no way to do so, by design — so we will generally need the camera owner's cooperation, and we will tell you honestly if we cannot locate any data about you.

4. How we use information

• Classify motion events at your outdoor cameras to notify you when a dangerous wild animal is detected, according to your per-species settings.
• Store timestamped sighting snapshots you can review in your history.
• Send you push notifications according to your tier and preferences.
• Maintain your account, including subscription state.
• Send you a sighting alert email for each sighting you are notified about. It is a service (transactional) notification about your own cameras, so it is on by default. It contains only the detection details — the animal, the camera name, the date and time, and the confidence score — plus a link to your dashboard. It contains no snapshot or other image: no image is embedded, attached, or linked, and the frame is viewable only when you are signed in to the app or web dashboard. You can turn it off at any time from your Account settings — in the web console or the Android app — or with the one-click unsubscribe link in every alert email; turning it off does not affect push notifications or your sighting history. At most one email is sent per notified sighting — there is no follow-up, escalation, or repeat.
• Send you, only if you opt in, a weekly email digest ("Your weekly wildlife report") summarizing the sightings recorded at your cameras over the previous 7 days (species, counts, and camera names — no snapshot images are attached). The digest is off by default; you can turn it on or off at any time in your Account settings (web console or Android app), and every digest email contains a one-click unsubscribe link that disables it immediately. A week with no sightings sends no email. Apart from the sighting alert email above, this opt-in digest, and sign-in one-time codes, we send no other email. We send no marketing email, and we never send SMS or place phone calls.
• Tune detection thresholds using only the explicit feedback you submit (e.g. marking an event "not a coyote" or "not interesting"). Ring media and data are not used to train or improve any AI/ML model; processing is inference-only.
• Keep the Service secure and available — verifying webhook signatures, preventing replay and abuse, throttling sign-in attempts, and diagnosing faults.
• Comply with legal obligations and respond to lawful requests.

We do not sell your personal information. We do not "share" personal information for cross-context behavioural advertising as defined under CCPA/CPRA.

4.1 Legal bases for processing (EEA and UK)

If you are in the EEA or the UK, the GDPR / UK GDPR requires us to tell you the legal basis for each purpose:

Where we rely on legitimate interests, you have the right to object — see §7.

5. How we share information

We do not share your snapshots with any third party for analysis. The animal-presence check described in §2.2.1 runs on Erbacci's own inference server — that is us, not an outside company, and it is not a disclosure of your images to anyone else. Beyond the parties listed below, no one receives your camera images.

• Amazon Web Services — infrastructure provider (storage, compute, vision model invocation, and the Amazon Rekognition fallback check described in §2.2.1).
• Hosting providers for our own inference server — our inference server reaches the internet through rented server capacity supplying its reverse proxy in Germany, and the GPU hardware itself sits in a facility in the United Arab Emirates with its own connectivity provider. These providers host the traffic and the hardware; they are not given your images to analyze and perform no processing on them, in the same way AWS hosts (but does not repurpose) the rest of the Service.
• Anthropic, via the Amazon Bedrock service, for vision model inference. Snapshots are not retained by the model provider beyond the inference call.
• Apple and Google, for push notification delivery. Android push is delivered through a Google Firebase project that Erbacci also uses for other Erbacci applications; what is sent there is the push token and the notification, never your images.
• Resend — transactional email provider, for delivery of sign-in one-time codes, the sighting alert emails, and — only if you opt in — the weekly wildlife report digest. Resend receives the recipient's email address and the message content (for an alert: species name, camera name, timestamp, and confidence score; for the digest: species names, sighting counts, and camera names); it does not receive Ring camera media files, snapshot images, or account tokens.
• Ring (Amazon), for OAuth identity, camera snapshots, and subscription webhooks.
• Law enforcement or regulators. We operate no law-enforcement data-sharing program and no law-enforcement integration. We disclose personal information to law enforcement or regulators only where compelled by valid legal process.

6. Retention

One consequence worth stating plainly. Snapshot images are kept for 90 days regardless of your tier, but the sighting record on the free/lapsed tier is kept for 30 days. So if you are not on the premium tier, a sighting can disappear from your timeline after 30 days while the underlying image remains in our storage for up to another 60 days before it expires. Deleting your account purges both immediately, at any time. We are reviewing this mismatch and intend to align the two.

7. Your rights

Depending on where you live, you may have the right to access, correct, delete, port, or restrict processing of your personal information. Exercise these rights in the web console at dangerousanimalalert.erbacciltd.com/app, or in the Android app (Account → Delete my data), or by emailing info@erbacciltd.com. We do not discriminate against you for exercising these rights.

7.1 If you are in the EEA or the UK

You have the following rights under the GDPR / UK GDPR:

• Access (Article 15) — a copy of the personal data we hold about you.
• Rectification (Article 16) — correction of inaccurate data. You can edit your camera and species settings yourself; write to us for anything else, including the email address on your account, which comes from Ring.
• Erasure (Article 17) — deletion of your account and data. See §7.3 for exactly what this does and what it does not reach.
• Restriction (Article 18) and objection (Article 21) — including the right to object at any time to processing we base on legitimate interests. There is no self-service control for these; write to us and we will action them.
• Portability (Article 20) — a machine-readable copy of the data you provided to us. We do not yet have a self-service export button, so we produce this manually on request.
• Withdraw consent (Article 7(3)) — for the weekly digest, at any time, from Account settings or the one-click unsubscribe link in any digest. Withdrawal does not affect processing carried out before it.
• Automated decision-making (Article 22) — we do not make decisions about you that produce legal or similarly significant effects. The Service classifies an animal and sends you a notification; the decision about what to do next is entirely yours.
• Complain to a supervisory authority. Our lead supervisory authority is the Office of the Commissioner for Personal Data Protection of the Republic of Cyprus, because Erbacci LTD is established in Cyprus. If you are in the UK you may complain to the Information Commissioner's Office (ICO). You may also complain to the authority where you live or work.

We respond to requests within one month, as Article 12(3) requires. If a request is complex or you have made several, we may extend that by up to two further months and will tell you why within the first month.

7.2 If you are in the United States

If you are a California resident, you have rights under the CCPA/CPRA including the right to know, delete, correct, and opt out of "sale" or "sharing" (we do neither). Residents of other US states with comprehensive privacy laws — including Virginia, Colorado, Connecticut and Utah — have comparable rights. Email info@erbacciltd.com and tell us which state you are in.

7.3 What deleting your account actually does — and what it does not reach

Deleting your account is a synchronous purge, not a scheduled expiry. In one operation we delete your push-notification device tokens and their push endpoints, every sighting record and every stored snapshot image, and finally your account record itself, including the Ring access and refresh tokens held on it. If any snapshot cannot be removed on the first attempt, we tell you so in the response rather than reporting a clean success.

Three honest limits:

• Backups. Encrypted point-in-time-recovery backups roll off after up to 35 days. A deletion does not reach inside that window; the data ages out of it.
• Logs. Diagnostic and access logs containing internal identifiers (and, for access logs, IP addresses) survive deletion and expire on the timers in §6.
• Your Ring authorization. Deleting your Dangerous Animal Alert account removes our copy of your Ring tokens, so we stop calling Ring on your behalf. It does not by itself withdraw the authorization you granted inside Ring. To remove that too, disconnect Dangerous Animal Alert in the Ring app (Menu → My Apps).

Separately, we have not verified what the inference server described in §2.2.1 writes to disk during a check, so we cannot state that a deletion reaches it. We would rather say that than imply a completeness we have not confirmed.

8. Security

• Encryption in transit (TLS) and at rest (AWS managed encryption). This includes the animal-presence check in §2.2.1: the image is transmitted only over TLS and only with a secret bearer token held in AWS Secrets Manager, and the request carries no information about you beyond the image itself.
• HMAC signature verification on all incoming webhooks.
• Short-lived JWT session tokens stored in iOS Keychain / Android Keystore.
• Least-privilege IAM policies on every backend component.
• Private, non-public storage for every snapshot, reachable only through short-lived signed links issued to the signed-in account holder.

Your Ring access and refresh tokens are held on your account record in our database, protected by the encryption-at-rest and least-privilege access controls above, and are deleted when you delete your account.

No system is perfectly secure. If we ever discover a breach affecting your data, we will notify you and the relevant authorities as required by law — including, where the GDPR / UK GDPR applies, the Cyprus Commissioner for Personal Data Protection or the ICO within 72 hours of becoming aware of it where the breach is notifiable.

9. Children

The Service is for adults. Our Terms of Service require you to be at least 18 years old and to be the owner of, or authorized to operate, the Ring cameras and Ring account you connect. We do not offer the Service to children and we do not knowingly create accounts for them.

• United States. The Service is not directed to children under 13, and we do not knowingly collect personal information from children under 13 (COPPA).
• EEA. Article 8 GDPR sets the age of consent for information-society services at between 13 and 16 depending on the Member State. Because account holders must be 18 or over, we do not rely on a child's consent for anything.
• United Kingdom. The Service is not designed for or directed to children, and the account holder must be an adult.

What we will not claim. We cannot say that no child's data is ever processed. An outdoor camera can capture a child as a passer-by or a visitor, and where the Amazon Rekognition fallback runs, a generic scene label such as "child" can be stored alongside the sighting (see §2.3). We do not identify that child, do not attempt to, and could not do so with the systems we run. If you believe we hold information about a child and you want it removed, contact us and we will delete it.

10. International transfers

Your account data and your stored snapshots are held in the United States, in Amazon Web Services region us-east-1. Species classification runs on Amazon Bedrock in AWS's United States regions (Bedrock may serve a request from any US region in the inference profile we use, so we do not claim it is pinned to a single one).

One processing step happens outside the United States, and it involves two countries. The brief animal-presence check described in §2.2.1 does not run in the US. The image is transmitted over an encrypted connection to a reverse proxy hosted on rented server capacity in Germany, and from there to a dedicated GPU inference server owned and operated by Erbacci LTD in the United Arab Emirates, where the yes/no check is performed. Your snapshot therefore transits Germany and is processed in the United Arab Emirates. Both facts are stated here because both are true: routing the image through Germany is itself a processing of it, and the analysis itself happens in the UAE.

What crosses the border, and what does not. Only the single still frame crosses, and only for the binary "is there an animal, yes or no" pre-filter. No name, email address, account identifier or camera name is sent with it. The species classification does not happen abroad — the animal is named by Amazon Bedrock inside AWS in the United States — and the snapshot you can review in the app is stored, and stays, in AWS us-east-1 in the United States.

If you are in the EEA or the UK. EU and UK data protection law applies to this processing — including because Erbacci LTD is established in Cyprus, an EU Member State. Two transfers out of the EEA/UK are involved, and we describe both.

1. Your account data and snapshots, to the United States. Everything we store — your account record, your sighting history and your snapshot images — is held in AWS in the United States, and the species classification runs there. This is a restricted transfer under Chapter V. Amazon Web Services is certified under the EU–US Data Privacy Framework (and its UK extension) and additionally offers Standard Contractual Clauses; those are the mechanisms relied on for the AWS leg.

2. The single frame, to the United Arab Emirates. The United Arab Emirates is not part of the EEA, and the European Commission has not adopted an adequacy decision for it (nor has the UK government issued UK adequacy regulations for it). Sending the frame there is therefore a restricted international transfer under Chapter V of the GDPR / UK GDPR. The German leg is inside the EEA and is not itself a restricted transfer.

We will not overstate the safeguards for that second transfer. An Article 46 transfer instrument for the UAE leg is being put in place and is not yet finalised. Because the receiving equipment is Erbacci's own rather than a separate company's, standard Standard Contractual Clauses cannot simply be signed between two counterparties, and we are taking legal advice on the correct instrument. Until that instrument is executed, we are not claiming to have one.

What is true today is that the following measures apply to that transfer: it carries the image and nothing else — no name, email, account identifier or camera name; it is encrypted in transit (TLS) and authenticated with a secret token; the only question asked is binary; no storage is requested; and the entire step can be switched off, in which case the frame never leaves AWS and the check is performed by Amazon Rekognition in the United States instead.

You can ask us about the current status of the transfer mechanism, and request details of the safeguards in place, by writing to info@erbacciltd.com. If you would prefer that your frames not be sent for that check, contact us.

If you are in the United States. This one step is still a transfer of your image outside the United States — to Germany in transit and to the United Arab Emirates for the check — and it is governed by the laws of those countries while it is there. We state it plainly rather than describe the Service as US-only processing, which it is not.

11. Changes to this policy

We may update this Privacy Policy. If changes are material, we will notify you in-app before they take effect.

12. Contact and controller details

Erbacci LTD, the data controller · info@erbacciltd.com
Registered office: Strovolou 77, Strovolos Center, Office 301, Strovolos 2018, Nicosia, Cyprus (Company No. HE 457237).

Erbacci LTD is established in Cyprus, an EU Member State. GDPR therefore applies to this processing by virtue of that establishment, wherever you live. Our lead supervisory authority is the Office of the Commissioner for Personal Data Protection of the Republic of Cyprus. Because we are established in the Union, no Article 27 EU representative is required.

Billing for the Dangerous Animal Alert subscription is handled by Erbacci LLC (Wyoming, USA) through the Ring Billing System, as described in the Terms of Service. Erbacci LTD, not Erbacci LLC, is the controller of the personal data described in this policy.