Privacy policy
Last updated
In short
- The houses you draw are stored in your browser, not with us. They only leave your device if you export them yourself or use the AI assistant.
- If you have an account, we store your name, email address, your plan and how much you have used the assistant. Conversation text is stored only when AI conversation review is enabled.
- When you type to the assistant, we send your messages, a summary of the house and any photos or drawings you attach to Anthropic. Normally your messages also pass through OpenAI’s realtime model, which answers first and tells you what the assistant is doing. In a live call, your voice goes straight to OpenAI, and what you ask normally goes on to Anthropic as text, with a summary of the house. Both say in their terms that they don’t train their models on this data.
- Stripe handles payments. We never see your card number.
- We count visits with our own simple statistics, without cookies and without being able to recognise you from one day to the next. We use no ads or third-party tracking, and the only cookies we use are the ones that make logging in work.
- You can download your data and delete your account yourself under Account.
- We also keep short-lived technical support diagnostics, linked to your account when you are signed in.
Who is responsible
MEE AS (organisation number 912 185 001) makes and runs House Madness. We are the controller of the personal data described here. This policy covers the website housemadness.app, the app and your account.
- Address: Svendsemarka 38, 3409 Tranby, Norway
- Privacy email: mail@meesoft.no
We have not appointed a data protection officer, because the law does not require one for a business like ours. Send any privacy question to the email address above.
Your drawings stay on your device
The houses you draw, and the PDFs and images you import, are stored in your browser (IndexedDB) on the device you use. The automatic tracing of scanned drawings also runs in your browser. We have no copy and cannot see your drawings. There are two exceptions: a drawing you attach in the chat with the assistant, which is sent to Anthropic as images (see “Photos and drawings”), and a zoning plan you have the assistant read in the permit dialog (see “Reading a zoning plan”).
Information about your house only leaves your device when
- you save a project file, print or export. You decide where the file goes.
- you use the AI assistant. See “How the AI assistant uses your data”.
- you look up an address or view maps. Your browser then fetches map data directly from public map services. See “Maps and addresses”.
Because your drawings are only with you, you also hold the backup. If you clear your browser data, the houses are gone. Use “Save project file” in the app to keep a copy.
What we process, why, and on what legal basis
The General Data Protection Regulation (GDPR) requires a legal basis for everything we do with personal data. Here is all of it, by purpose. How long we keep each kind of data is under “How long we keep data”.
Your account
You don’t need an account to draw. You need one to buy a plan and to use the AI assistant.
- What: your name, email address, whether the address is confirmed, and when the account was created and changed. Your password is stored only as a cryptographic hash (scrypt), so we cannot read it.
- Country preference: we store the country you choose as the default for new projects. You can change it in your account, and choose another country for an individual project.
- Logging in with Google, Apple or Vipps: if you choose this, we receive your name, email address, an ID for your account with that service and the login tokens it gives us. From Google we also receive a link to your profile picture.
- Why: to create your account, log you in on the website and in the app, and give you what you have bought.
- Legal basis: our contract with you (GDPR Art. 6(1)(b)).
We need your name and email address to create an account. Without them you cannot have an account, but you can still draw.
Payments
- What: your Stripe customer ID, what you have bought (subscription, Project pass or top-up), its status and dates, whether you pay the founder price, and your AI minutes. Stripe processes your card details, billing address and country (to work out the right VAT), and VAT numbers for businesses. Your card number never reaches us.
- Where from: from you when you pay in Stripe’s checkout, and from Stripe.
- Why: to take payment, give you access, send receipts, work out VAT, keep accounts, and handle refunds and complaints.
- Legal basis: the contract (Art. 6(1)(b)) and our obligations under the Norwegian Bookkeeping Act and VAT rules (Art. 6(1)(c)).
Stripe also uses some data for its own purposes, such as preventing fraud and meeting anti-money-laundering rules. Stripe is responsible for that itself. Read Stripe’s privacy policy.
The AI assistant
- What: your messages, photos and drawings you attach, zoning plans you have read, a summary of the house and the results of the tools the assistant uses. In a live call, also your voice and a written version of what is said. We track how much you have used: the number of questions, rounds, tokens and minutes per month. We also keep fingerprints of what the assistant itself writes and says; see below.
- Why: to give you answers and suggestions, keep track of the minutes and limits in your plan, and prevent misuse.
- Legal basis: the contract (Art. 6(1)(b)). The limits that prevent misuse and keep costs down rest on our legitimate interest in keeping the service available and safe for everyone (Art. 6(1)(f)).
More under “How the AI assistant uses your data”.
AI conversation review
We may keep conversation text for up to 30 days to improve the assistant. Authorised administrators may review your messages, the assistant’s replies and written transcripts of live calls. These records are linked to your account and included in your data export. We remove them from active storage when you delete your account. Audio and attachments are excluded.
Our legal basis is our legitimate interest in improving the assistant (Art. 6(1)(f)). You can object as described under ‘Your rights’.
For client data in Proff accounts, the data processing agreement in the terms applies.
Email from us
- What: your email address and name, and the content of the email.
- Why: to confirm your email address, send password reset links and send you important messages about your account and contract. These can be that a subscription is running and how to cancel it (the law requires this at least every six months), that your Project pass will end soon, that you have withdrawn from a purchase, or that the terms are changing. Receipts are sent by Stripe.
- Legal basis: the contract (Art. 6(1)(b)). For notices the law requires: legal obligation (Art. 6(1)(c)).
We don’t send newsletters or marketing. If we want to later, we will ask for your consent first.
When you contact us
- What: what you write, your email address and what we need to help you, such as which purchase your question is about.
- Why: to answer you, solve the matter and document what we agreed.
- Legal basis: the contract (Art. 6(1)(b)) when the matter concerns it. Otherwise our legitimate interest in answering enquiries (Art. 6(1)(f)).
Security and logs
- What: for each login we store the IP address, browser and device (user agent) and time. The server and our hosting provider log technical data such as IP addresses, times, the address requested and error messages.
- Why: to keep your account and the service secure, detect and stop attacks and misuse, and fix errors.
- Legal basis: our legitimate interest in protecting accounts and the service (Art. 6(1)(f)), and our duty to keep data secure (Art. 32).
Limiting attempts
- What: your IP address and how many attempts come from it, for example to log in, sign up or reset a password. For the assistant and live calls we count requests and calls per account.
- Why: to stop password guessing, spam and runaway requests.
- Legal basis: our legitimate interest in protecting the service and its users (Art. 6(1)(f)). The counters are kept only in the server’s memory and reset after an hour at most.
Technical support diagnostics
We keep an operational support timeline separate from anonymous statistics: operation names (such as opening Drawings or printing), technical error messages, HTTP status codes, code locations, the app and server version, and a coarse browser and device description. When you are signed in, these records are linked to your account so we can investigate problems and provide support. A temporary page-session code stays only in memory, not browser storage.
These technical diagnostics do not record keystrokes, form values, screenshots, drawing files or conversations. Credentials, URL query strings and quoted values are removed from diagnostic messages. Only authorised administrators can read the records on our own server.
These records are kept for up to 30 days, or less during high traffic, and are included in your data export and removed when you delete your account. Signed-in operational browser reports are collected for every account, including when Global Privacy Control or Do Not Track is set. The purpose and legal basis are the same as for fixing errors under Security and logs.
Interaction diagnostics
House Madness automatically records cursor movements, clicks, scrolling, interface layout and technical error markers for signed-in accounts to diagnose problems. Drawings, uploads, chat messages, passwords and typed values are excluded.
Diagnostics are linked to your account, kept on our own server for up to 7 days and accessible only to administrators. This operational collection applies to all existing and new signed-in accounts and cannot be disabled in the app.
Authorised administrators can review a text journey and a structural replay. The purpose and legal basis are the same as for error diagnostics described above. Records are included in your data export and removed when your account is deleted. Temporary page identifiers and the upload buffer stay in memory; no additional cookies or browser storage are created. Earlier diagnostics preferences, Global Privacy Control and Do Not Track do not disable this account-linked operational collection. Anonymous statistics remain separate and continue to honour those browser signals.
Statistics
- What: for each page you see on the website or in the app, and a few steps such as opening the app, drawing rooms, opening the drawings, printing, using a calculator, signing up, going to checkout or writing to us: the time, the page’s address without anything personal in it, the site you came from (its name only) and the campaign in the link, your country, the language, whether you use a phone, a tablet or a computer, and a code made from your IP address and browser with a key that changes every day. We don’t store the IP address or the browser, we can’t tell who you are, and we can’t recognise you from one day to the next. Nothing about your house is counted, and nothing is tied to your account.
- Why: to know how many use the website and the app, where they come from and where they give up, so we can make the service better.
- Legal basis: our legitimate interest in understanding and improving the service (Art. 6(1)(f)). If your browser sends Global Privacy Control or Do Not Track, we don’t count you at all.
Marking what the assistant writes and says
The EU’s AI Act (Article 50) requires that what an AI writes or says is marked as made by AI, in a way machines can read.
- What: each answer the assistant writes in the chat, each note it writes on your drawings, the texts it writes for a revision and the written version of what it says in a live call gets a marking signed by us: that it was made by AI, by which model and when, with a fingerprint (a hash) of the text. The marking stays with the text in the app, when you copy it, and in the project file. We also keep fingerprints of these texts: hashes of their word sequences.
- Why: so that anyone can check whether a text came from our assistant. On the AI checker, anyone can paste a text and see whether it did. What is pasted there is not kept.
- Legal basis: our legal obligation under the AI Act (Art. 6(1)(c)).
Maps and addresses
When you look up an address or view maps and aerial photos in the app, your browser fetches the data directly from these services, not through us:
- Kartverket and Geonorge (the Norwegian mapping authority): address search, property, elevation data and maps
- PDOK (the Dutch government’s geodata service, run by the Kadaster): address search, parcels, aerial photos and elevation data in the Netherlands
- The IGN and the CNIG (Spain’s national mapping agency): address search (CartoCiudad), aerial photos (PNOA) and elevation data in Spain
- The Dirección General del Catastro (Spain’s land register): parcels in Spain
- Esri: aerial photos outside the Netherlands and Spain, and there when PDOK or the IGN does not answer (servers in the United States)
- Google (Google Maps Platform): the address search with Google’s suggestions, in countries without a public address search and where the public one finds nothing. When you use it, your browser loads Google’s search from Google and sends Google what you type and your IP address. Google Ireland Limited is responsible for this as an independent controller, under Google’s privacy policy. Google may process the data in the United States, under the EU–US Data Privacy Framework. The app keeps only Google’s ID for the place you pick, in your project.
- Overpass, a service for OpenStreetMap data: buildings in the neighbourhood (run by FOSSGIS e.V. in Germany)
- Statens vegvesen (the Norwegian Public Roads Administration): the roads round the house in Norway, from its road database (NVDB)
- The Norwegian Building Authority (Direktoratet for byggkvalitet, DiBK): the zoning plans at the house in Norway, from its national plan base (Nasjonal arealplanbase)
These services see your IP address and what you search for or the area you look at, just as every website sees the IP address of its visitors. We don’t receive this information. We chose these services, and they are used only when you use the map features (the contract, Art. 6(1)(b)). Each service is responsible for what it does with the data.
How the AI assistant uses your data
The AI assistant is artificial intelligence (AI), not a person. You can type to it, attach photos and drawings, or talk to it in a live call.
The assistant is optional. All the drawing tools, the 3D view and the calculations work without it, and nothing is sent to the AI providers until you click in the assistant’s message field, send a message or start a call yourself.
When you type
The answers about your house are written by Anthropic’s Claude language model. Each time you send a question, this goes through our server to Anthropic:
- your message and the whole conversation so far. The model remembers nothing between questions, so the conversation is sent again each time.
- a short summary of the house: the project name, the floors and rooms, and which floor you are looking at
- the results of the tools the assistant uses, such as the sizes of rooms, walls and windows, loads and beams, and the address of the property if you have entered it
- the photos and drawings you have attached in the conversation
- a written version of what was said in live calls, if you have talked to the assistant along the way
Our server adds instructions and descriptions of the tools and sends the answer back to you. The chat in the app resets when you close the app or start a new conversation. See AI conversation review.
Normally, OpenAI’s realtime model also takes part when you type. When you click in the message field, the app connects your browser directly to OpenAI (text only, without the microphone), and our server starts the session with instructions, the app’s language and the latest part of the conversation (text only, no photos or drawings). Each message you type goes over this connection to OpenAI’s model. It answers small talk, such as a greeting, itself. Anything else it passes on to Claude, and while Claude works it writes a short line about what is happening. For this it receives short notes on what Claude is doing, and Claude’s answers. Photos and drawings are not sent to OpenAI; it is only told that something is attached. Without the realtime model, what you type goes only to Anthropic.
Photos and drawings
You can attach photos, for example of a ceiling to find out whether there is a beam. Before a photo is sent, the app makes it smaller (at most 1568 pixels on the longest side) and saves it again as a JPEG. This also removes hidden information in the file, such as the GPS position. You can attach up to 3 photos per question and 6 per conversation. The photos are sent again each time, for as long as the conversation lasts.
You can also attach your drawings, as a PDF or a picture, so the assistant can draw up the house from them. The app sends images of the pages, made smaller in the same way, never the file itself. Each page counts as a photo towards the limits above, and when a PDF has more pages than fit, one overview of its pages is sent instead. The tracing of the drawings into walls runs in your browser. The assistant then gets the result as numbers, such as the rooms, their sizes and the scale, and, when you type, an image of each traced page with the trace drawn on it. Drawings often show the owner’s name, the address and the property numbers in the title block, and the assistant then sees those too.
Please avoid photos of people, of documents other than your drawings, or of anything else the assistant doesn’t need.
Reading a zoning plan
In Norway you can upload a zoning plan’s provisions and map in the permit dialog (“Do I need to apply?”) and have the assistant read the plan’s limits. The files you choose (up to 3 PDFs or pictures) go through our server to Anthropic as they are, with the address and the property numbers (gnr/bnr) if you have entered them, so that the right zone can be found. We don’t store the files or the answer. The values are filled in only when you confirm them, and are then kept with the house on your device. Each reading counts as one question towards your plan’s limits.
In a live call
Live calls use OpenAI’s Realtime API. When you press “Talk live” and allow your browser to use the microphone:
- The sound from your microphone goes straight from your browser to OpenAI (over WebRTC), and the answer comes back as speech. The audio does not pass through our server.
- If you typed with the assistant before the call, the app sends OpenAI, through our server, the latest part of the typed conversation (text only, no photos or drawings), so you can carry on where you left off.
- OpenAI writes down what you say and what the assistant answers. This transcript is shown in the chat.
- Normally, OpenAI’s model passes what you ask on to Claude as text and reads Claude’s answers out to you. Our server then sends Anthropic the same as when you type, including a summary of the house. OpenAI receives short notes on what Claude is doing, Claude’s answers and anything you type during the call.
- If Claude is not used in the call, OpenAI’s model answers on its own. OpenAI then also receives a summary of the house and the results of the tools the assistant uses, as Anthropic does when you type.
Our server starts the call, notices when it ends, and deducts the minutes from your account. Our server does not store call audio. The microphone is only in use during the call, and you can turn it off at any time.
The app also has a simpler voice option, “Talk (preview)”, which uses your browser’s own speech recognition. In some browsers, such as Chrome, the browser then sends the audio to the browser maker to turn it into text, and the text of the answer may be sent there to be read aloud. That happens under your browser’s own terms, and we don’t receive the audio.
Do they train on your data?
- Anthropic says in its Commercial Terms that it does not train its models on content sent through its API. According to Anthropic, inputs and outputs are normally deleted within 30 days, but may be kept longer when needed to investigate misuse or when the law requires it. See Anthropic’s page on retention.
- OpenAI says in its documentation on API data that data sent to its API is not used to train its models unless the customer has chosen to allow it. We have not. OpenAI normally keeps logs for detecting misuse for up to 30 days.
This is the providers’ own description of how they handle data. We have agreements with both on how they must handle the data, but we cannot inspect their systems ourselves. We don’t use your conversations to train AI models either.
The answers can be wrong
The assistant can misunderstand, miscalculate or misread a photo or a drawing. Suggestions about the load-bearing structure are rough estimates, not structural design, and must be checked by a qualified structural engineer (in Norway, a RIB). See the terms for more.
No automated decisions
The assistant makes suggestions, and you decide whether to accept or discard them. We make no decisions about you based solely on automated processing that have legal effects or similarly significantly affect you (GDPR Art. 22).
If you use House Madness for work
If you use Proff in work for clients and send information about them to the assistant, such as their address, we process that information on behalf of your business. Your business is then the controller of the client data, and we are the processor. The data processing agreement is in the terms, under “Business customers (Proff)”.
Who we share data with
We never sell personal data. We share it only with providers we need to run the service. A processor handles data only on our instructions and under a data processing agreement with us.
| Who | What they do | Role | Where | Basis for transfers outside the EEA |
|---|---|---|---|---|
| Amazon Web Services EMEA SARL (Luxembourg) | Runs the server and the database (AWS Lightsail, in a data centre in the EU) | Processor | EU | EU standard contractual clauses in AWS’s data processing agreement, for any access from outside the EEA |
| Cloudflare, Inc. (United States) | Sits in front of the website and the app: delivers the pages, encrypts the connection and stops attacks. It sees every request, with your IP address | Processor | Its data centres worldwide (the one nearest you answers) and the United States | EU–US Data Privacy Framework and EU standard contractual clauses |
| Stripe (Stripe Payments Europe, Ltd., Ireland) | Payments, receipts, VAT and the customer portal | Processor. Independent controller for its own purposes, such as preventing fraud | EU and United States | EU–US Data Privacy Framework and EU standard contractual clauses |
| Anthropic (Anthropic Ireland, Limited) | The assistant’s language model (Claude), for typed questions and normally also for what you ask in a live call | Processor | United States | EU standard contractual clauses |
| OpenAI (OpenAI Ireland Ltd) | Live calls, and normally the realtime model that passes on what you type | Processor | United States | EU standard contractual clauses |
| Mailgun Technologies, Inc. | Sends emails about your account and your orders, from its EU region | Processor | EU (support from the United States) | EU standard contractual clauses |
| “Log in with Google”, if you choose it, and the address search with Google, where it is used | Independent controller | – | – | |
| Apple | “Log in with Apple”, if you choose it | Independent controller | – | – |
| Vipps MobilePay AS (Norway) | “Log in with Vipps”, if you choose it | Independent controller | Norway | – |
Google, Apple and Vipps handle your data under their own privacy policies when you log in with them. We only receive what is listed under “Your account”.
We may also share data with our accountant and auditor, with lawyers if we need to establish or defend a legal claim, and with public authorities when the law requires it.
Transfers outside the EEA
Some of our providers process data in the United States. For those transfers we use a lawful basis under Chapter V of the GDPR:
- The EU–US Data Privacy Framework for companies certified under it. The European Commission’s decision on the framework also applies in Norway through the EEA Agreement.
- The EU standard contractual clauses (SCCs) in the data processing agreements, including those with Anthropic and OpenAI.
On top of this, data is encrypted in transit, and we send no more than is needed. You can ask for a copy of the standard contractual clauses. Trade secrets may be redacted.
How long we keep data
| Data | How long |
|---|---|
| Interaction diagnostics | Up to 7 days |
| Your account (name, email, login methods) | Until you delete your account. If you haven’t logged in for three years and have neither an active plan nor bought AI minutes left, we delete the account. We email you at least 30 days before. |
| Login sessions (IP address, browser) | Until you log out, or until the session expires 30 days after you last used it |
| Links to confirm your email or set a new password | Valid for one hour. Password reset links are deleted once used or expired. |
| Plan, pass, AI minutes and assistant usage | For as long as you have an account |
| Fingerprints of what the assistant writes and says | 12 months |
| Payments and accounting records | Five years after the end of the financial year, as the Norwegian Bookkeeping Act requires. Records of sales to consumers in the EU: ten years, as EU VAT rules require. |
| Your request for the service to start straight away (right of withdrawal) | For as long as the contract lasts, and three years after it ends |
| Messages to us | Three years after the matter is closed |
| Technical logs | Up to 30 days |
| Counters for limiting attempts | Up to one hour, in memory only |
| Statistics (pages seen and steps, without your IP address) | 13 months. The daily key is deleted after two days. |
| Backups | Data we have deleted disappears from backups within 30 days |
| The content of conversations with the assistant | Up to 30 days when AI conversation review is enabled. At Anthropic and OpenAI: see “Do they train on your data?” |
Three years is the general limitation period for claims under Norwegian law. For that long we may need to document what we agreed.
Your rights
You have the right to
- know what data we hold about you and get a copy of it (access, Art. 15)
- have incorrect data corrected (Art. 16)
- have data erased (Art. 17)
- ask us to restrict processing, for example while we check whether something is wrong (Art. 18)
- receive data you have given us in a machine-readable format (data portability, Art. 20)
You have the right to object. Where we process data on the basis of legitimate interest (security, logs, limiting attempts, statistics and limits against misuse), you can object to the processing on grounds relating to your particular situation (Art. 21). We will then stop, unless we have compelling legitimate grounds that override your interests, or need the data to establish, exercise or defend a legal claim.
Do it yourself under Account
- Download my data gives you a file (JSON) with your account, plan and AI usage. If you also want your login history or payment details, write to us.
- Your name can be changed under Account. To change your email address, write to us.
- Delete your account deletes the account straight away. If you have only logged in with Google, Apple or Vipps and have no password, write to us and we will delete it for you.
When your account is deleted:
- we delete the account, your login sessions, the links to Google, Apple and Vipps, your plan, your AI minutes, your usage data and the link to your customer record at Stripe
- active subscriptions end straight away. The terms explain what this means for the rest of the period and for unused minutes.
- we and Stripe keep payment and accounting records for as long as the law requires
- the data disappears from logs and backups within 30 days
- the drawings in your browser stay where they are
Write to us
Email mail@meesoft.no. It’s free. We answer within one month. If the request is complicated, we may extend this by up to two more months, but we will tell you within the first month. To protect your account, we may ask you to confirm that it’s you, for example by replying from the email address on the account.
You can complain
If you think we process personal data in breach of the rules, you can complain to the Norwegian Data Protection Authority, Datatilsynet. If you live in another EU or EEA country, you can also complain to the data protection authority there. See the list of data protection authorities in the EEA. You are welcome to contact us first. We can often sort things out quickly.
Cookies and local storage
We only use cookies and local storage that are strictly necessary for the service you have asked for. Under the Norwegian Electronic Communications Act (ekomloven), these need no consent. Our statistics use no cookies and store nothing on your device, and we use no ads or third-party tracking, so there is no consent banner. If we ever want to use something that needs consent, we will ask you first.
Cookies
| Name | What it does | How long |
|---|---|---|
__Secure-hm.session_token |
Keeps you logged in on the website and in the app | 30 days, renewed when you use the service. If you untick “Remember me on this device”, it is deleted when you close your browser. |
__Secure-hm.dont_remember |
Remembers that you don’t want to stay logged in | Until you close your browser |
__Secure-hm.state |
Protects logging in with Google, Apple or Vipps against forgery | 5 minutes |
Local storage in your browser
| Name | What it does |
|---|---|
IndexedDB (project:…, projects:index, blob:…) |
Your houses and the files you have imported |
Cache Storage (hm-app-…), kept by the app’s service worker |
The app’s own files (its code, texts and pictures), so that it starts without the internet. Nothing of yours is in them |
housemadness.editor |
The app’s settings, such as language and colour theme |
housemadness.lastProject |
Which house you had open last |
hm-theme |
The website’s colour theme: light or dark |
hm-user |
Your first name, so the menu shows at once that you are logged in. Deleted when you log out. |
This stays in your browser and is not sent to us. You can delete it in your browser’s settings. That also deletes your houses, so save a project file first.
Children
House Madness is not made for children under 16. You must be at least 16 to create an account, and at least 18 to buy anything and to use the AI assistant. If we learn that a child under 16 has created an account, we will delete it. If you are a parent and think your child has given us personal data, write to mail@meesoft.no.
Security
This is how we protect your data:
- All traffic is encrypted (HTTPS).
- Passwords are stored only as a hash.
- The login cookie cannot be read by scripts on the page (HttpOnly) and is only sent over an encrypted connection.
- We check where requests come from and stop repeated login attempts.
- The server and database are in the EU/EEA, and backups are protected.
- Only the people who need it to run the service have access to the data.
If a personal data breach happens anyway, we report it to Datatilsynet within 72 hours when the law requires it, and tell you if the breach is likely to put you at high risk.
Changes to this policy
We update this policy when the service or the rules change, for example when we start using a new provider. The date at the top shows when it last changed. We show the AI conversation review notice in the app before collecting conversation text. For other changes that matter to you, we will tell you by email before they apply. You can get earlier versions by writing to us.
Contact
MEE AS
Svendsemarka 38, 3409 Tranby, Norway
Privacy email: mail@meesoft.no