Privacy in plain language
Data and AI in plain language
Where your data goes, when and what you can turn off. A short summary; the privacy policy remains authoritative. Read the privacy policy
The essentials in five sentences
1. Your beekeeping data is stored on Hetzner servers in Germany.
2. Recommendations, alerts and the daily briefing run on our server using fixed rules, without external AI.
3. Chat, photo findings, voice input and the AI features in “My presence” use AI with your consent; an AI checks your own free text in bloom reports for abusive content even without this consent.
4. The necessary content goes to Anthropic, including to the USA; AssemblyAI, a US company with an EU endpoint, transcribes voice recordings.
5. We do not sell personal data or report a suspected bee disease to authorities ourselves; fulfilling the notification duty is your responsibility as the beekeeper.
Where each type of data goes
All recipients outside HiveKraft, listed by the data they receive.
| What | To whom | Where | When | Turn off |
|---|---|---|---|---|
| Account, colonies, entries and photos | Hetzner | Germany | To run HiveKraft | Required to use HiveKraft |
| Question, conversation, profile and colony data extract | Anthropic (contracting party: Anthropic Ireland) | Including USA and other countries | When you use AI chat | Withdraw AI consent |
| Photo, type and prompt; inspection photos also include strength and season | Anthropic | Including USA and other countries | During photo analysis; with consent, also automatically for inspection photos | Withdraw photo consent |
| Audio and the names of your active colonies and your apiaries | AssemblyAI | US company, EU endpoint; US access cannot be ruled out | For voice input, rounds and dictation buttons | Withdraw voice consent |
| Voice input transcript; for a round, colony numbers and names and, per colony, hive type, last inspection and treatment | Anthropic | Including USA and other countries | Then for structured analysis | Through voice consent |
| Selected page content, messages, settings or approved image previews | Anthropic | Including USA and other countries | In “My presence”, with consent for each AI feature | Withdraw consent for the feature; design without AI |
| Custom plant name or note in a bloom report | Anthropic | Including USA and other countries | For eligible free text, also without AI consent | Choose a listed plant without a note; objection: privacy policy §9.5 |
| Rounded apiary position | Bright Sky (German Weather Service data)Request from our server | No processing location stated in the privacy policy | For weather, forecasts, alerts and the growing degree sum | No separate control for these requests |
| Rounded apiary position | Open-MeteoRequest from our server | Provider based in Switzerland; no processing location stated in the privacy policy | For fallback weather requests, elevation and the harvest forecast weather factor | No separate control for these requests |
| Search text or exact coordinates of a map point | OpenStreetMap Foundation (Nominatim)Request from our server | United Kingdom, Netherlands | When you start an address search or tap “Look up address for this map point” | Do not use address search |
| Channel number and read key of your hive scale, requested time range | ThingSpeak (MathWorks)Request from our server | Provider based in the USA | When checking the connection and then at the interval you chose (15 minutes to daily) | Delete the device or use another data source |
| Payment data and billing address | Stripe | USA | When you purchase a paid plan | Required for the purchase |
| Email address, name and the content of each email, such as links, codes, contract and colony details; account ID in unsubscribe links | Brevo | EU | With every email from HiveKraft, including your presence’s newsletter and visitor inquiries | Required for account and contract emails; turn off notifications and marketing emails in settings |
| Technical error data, browser and URL; IP addresses are removed | Sentry | USA; initial processing in the EU | When an error is reported | No separate control |
| From Google to HiveKraft: your name, your email address and whether Google has verified it, your Google ID and, where Google provides it, the address of your profile picture; we store the sign-in tokens only in encrypted form and do not use them for further access to Google | Google OAuth (Google is a controller in its own right, not our processor) | No processing location stated in the privacy policy | Only when signing in with Google | Sign in with email and password |
| Technical access metadata (IP address, user agent, requested URL, status code); CrowdSec reports IP addresses that repeatedly exhibit malicious behavior to a community reputation network | CrowdSec | EU, France | To defend against automated attacks | Required to protect the service |
| IP address and visible map area | OpenStreetMap Foundation (map tiles)Directly via your device or browser | Worldwide cache network | When loading a map | Do not open the map view |
| Exact coordinates, including all stops in the route planner | Google MapsDirectly via your device or browser | No processing location stated in the privacy policy | Via “Route”, a coordinate link or “Open in maps” | Do not open external map links |
| Browser endpoint and sending time; content encrypted for your device | Google, Mozilla, Apple or Microsoft (browser push service)Subscription via your device, sending via our server | USA | When you enable push and a notification is sent | Turn off push in settings |
| Audio during fallback dictation | Browser provider (Google or Apple)Directly via your device or browser | USA | For dictation buttons when AssemblyAI is unavailable | Do not use dictation; type instead |
We currently do not measure usage of this website.
What the AI receives and what it does not
What the AI receives
Your question, the conversation so far and the app page you are viewing.
An extract from your profile and records:
More
beekeeping experience, number of colonies and apiaries, colonies with name, hive type and notes, recent inspections, treatments, feedings and harvests, open tasks, moves, active withdrawal periods, current weather with temperature and conditions for up to five apiaries with their names, 7-day weather forecast (without the apiary’s name) and the phenology status of the apiary with the most colonies; in addition, the chat can run up to three lookups per message and transmit their results, for example apiaries (name and number of colonies, without coordinates; the address you entered only if you ask for it in the chat), queens, inventory and finance entries with amount and description. Hive-scale and sensor data (overview, alerts, device status and readings) only if you have switched on “Include IoT & hive scale”; readings that already appear in earlier replies of a conversation continue to be sent with the conversation history after you switch it off.
The photo, photo type and analysis prompt; for inspection photos, the recorded colony strength and season, and in chat also your message.
The transcript of your voice input for structured analysis; for a round, also the numbers and names of your active colonies and, for each assigned colony, hive type, last inspection and last treatment with date.
In “My presence”, the selected page content, messages and settings for the feature, but no image files; for an image description, the reduced preview image without file name and metadata, only for individually approved images.
What we do not add – and how the provider uses data
- We do not automatically add stored apiary coordinates to chat; your own text and older conversations may contain coordinates.
- Anthropic does not use the transmitted content to train its models.
- For bloom report moderation, we transmit only the free text, without adding stored location or profile data.
Data is not anonymized for AI processing. Personal information in your text or photos may also be transmitted. An AI photo finding is an indication, not a veterinary diagnosis.
How long data stays
- At Anthropic
According to its own information, Anthropic deletes inputs and outputs within 30 days.
More
Anthropic only keeps them longer if its safety systems suspect a violation of Anthropic's usage policy (content up to two years, classification scores up to seven years) or if the law requires it. There is no agreement on processing without any storage at the provider (zero data retention).
- At AssemblyAI
A retention period of one hour is set for our account (provider confirmation of 1 June 2026).
More
Deletion then begins; according to the provider, completion may be delayed by hours to days. After retrieving the transcript, we instruct AssemblyAI to delete the transcript and audio.
- With us
Account data (name, email, profile): until account deletion.
More
Consent records then stay linked to the pseudonymized account record; their retention period is set out in §7. Audio files from voice input are automatically deleted on our server after successful processing. Beekeeping data, chat histories, consent records and invoices have different retention periods, described with each feature and in section 14 of the privacy policy; retention periods at recipients are set out in sections 8 and 10. If you delete data with us or withdraw consent, this does not cover requests Anthropic has already received; the periods stated apply to them.
Consent and turning features off
Chat, voice input and photo analysis require your consent. You can withdraw each consent in the AI settings. Hive records, treatments, harvests, exports and rule-based recommendations can be used without AI.
The AI features in “My presence” ask for your consent separately.
More
You can also design your site without AI. Bloom free-text moderation is separate: no AI check takes place without eligible free text. Section 9.5 of the privacy policy describes how to object. Manage your AI consents here. You need to sign in.