Privacy and cookies

Dlubal Software GmbH, Am Zellweg 2, 93464 Tiefenbach, Germany, is responsible for processing your personal data as controller under the GDPR. For privacy questions or requests to access, correct, or delete your data, or to object to processing, contact our data protection officer at [email protected]. The responsible supervisory authority is the Bavarian State Office for Data Protection Supervision (BayLDA) in Ansbach.

The legal basis depends on why we process your data. Article 6(1)(b) GDPR applies when we need data to perform a contract with you or take steps you request before entering a contract. For legal obligations, we rely on Article 6(1)(c). When we rely on legitimate interests under Article 6(1)(f), we explain the interest in the relevant section. Features that require your consent rely on Article 6(1)(a).

Some account, sign-in, and billing details are necessary to provide those features. Without them, we may be unable to offer you an account, access, or a paid contract. You choose what to enter in a chat and which files to upload. CALC may be unable to complete a task without the information it needs. You do not have to turn on optional analytics or voice features.

Accepting optional cookies does not give blanket consent to all AI processing or international transfers. If you submit data about other people, processing their data also needs a legal basis. Our contract with you alone does not justify every use of their data.

When you visit CALC, we process the connection data needed to deliver the website. This includes your IP address, the time of access, the requested address, browser and operating system details, and technical status messages. We use this data to display the website, keep it secure, and investigate faults or abuse. The legal basis is our legitimate interest in a secure and available application under Article 6(1)(f) GDPR.

We run the application, databases, file storage, and backups using our own operational processes and infrastructure providers. A hosting provider may receive data or process it on our behalf even when we manage the software ourselves.

To create and manage your account, we process details such as your email address, name or display name, account ID, and sign-in information. We also process your language, preferences, and, where applicable, company and license details. If you sign in with an email address and password, we process the credentials needed to verify and secure your access.

Where offered, you can also sign in through Microsoft, Google, or your Dlubal account. The service you choose handles the sign-in. CALC receives the identity and profile details needed to match you to your account. The service's own privacy notice also applies to any processing it carries out for its own purposes.

We use this data for secure sign-in, account management, and the agreed features. The legal bases are Article 6(1)(b) GDPR and, for additional security measures, Article 6(1)(f).

To handle your request, CALC processes your input, the conversation context needed for a response, generated answers, your selected preferences, and results from the tools it uses. Depending on the task, this also includes uploaded files, images, drawings, model information, or content extracted from them. This content may be personal or confidential even if it contains no names.

We use external services for AI processing. Each provider receives the content CALC assembles for that step. Earlier messages and tool results may be sent again when you ask a follow-up question. The providers involved depend on the task, the model's capabilities, and the configured routing. A model name alone therefore does not tell you everyone who receives data.

We process this data to answer your request, perform calculations, provide explanations, or create documents. Article 6(1)(b) GDPR applies to a contracting individual's personal data that we need for this service. If project documents contain other people's data, the legal basis also depends on who is responsible for processing it and what the task involves.

Later analysis for product improvement serves a different purpose, so we describe it in its own section. That analysis excludes original uploaded files. Files may still be sent to an AI provider when CALC handles your original request.

An AI provider may process data to generate an answer, store it temporarily, or use it to train and improve a model. These activities need to be considered separately. Even when model training is excluded, data may still be transmitted or stored, for example in technical caches, for security checks, or to meet legal retention requirements.

The rules for storage and training depend on the service, business account, contract, and enabled features. A provider's statements about another product or plan therefore do not necessarily apply to CALC.

International providers may process data outside the EU and the European Economic Area or access it from those locations. A company's registered office does not necessarily tell you where processing takes place. Remote access and other providers it engages also need to be considered.

These transfers must meet Articles 44 et seq. GDPR as well as having a legal basis for the processing itself. Depending on the recipient, an applicable adequacy decision or appropriate safeguards may be available, such as standard contractual clauses with any necessary additional protection. Even with standard contractual clauses, the actual level of protection must be assessed for the specific transfer.

To ask which recipients and safeguards apply to your data, email [email protected].

The GDPR does not generally require AI models to run on privately operated GPU servers. What matters is how the data is processed. This requires a legal basis, clear information, defined responsibilities, and appropriate security. Transfers to third countries must also meet the applicable requirements.

The German supervisory authorities' guidance on artificial intelligence and data protection of May 6, 2024 covers internally operated and external AI applications. Section 2.1 explains when an external provider processes data on behalf of another organization. Section 1.7 describes technically closed systems as preferable for data protection.

External AI services are therefore not prohibited as a rule. Whether a particular cloud service is lawful still needs to be assessed individually. The guidance explains legal requirements; it is not a regulatory assessment or certification of CALC.

CALC can use preferences you provide, such as your language, preferred design standards, or response style. If you turn on memories, these details may be saved after a chat ends and used in later requests. We may send relevant memories to the AI provider along with a request.

Memories are stored separately from your chat history. You can view, edit, and delete them in settings. Turning the feature off does not delete memories already saved. The section on AI analysis of your chats explains product improvement analysis and when mood analysis takes place.

When CALC searches the web for your task, it sends search terms to Brave Search. Those terms may contain details from your task, so a search request is not automatically anonymous. When CALC retrieves an external page, its operator receives technical connection data and the requested address.

If you connect RFEM, a Connector, or other tools to CALC, the requested feature processes information such as connection details, tool calls, selected model or project data, and results. Results may enter the chat and be included in later AI requests. Even when a calculation runs on your computer, not all related data necessarily stays there.

Which data is sent to which recipient depends on the connected service and the action. For services you connect yourself, please also read the operator's privacy notice.

Sharing or import links allow others to view or import the content you share. What a recipient can do depends on the feature you choose. Keep these links confidential: depending on the sharing method, people who receive a forwarded link may also gain access. Revoking a link later does not automatically delete copies that recipients have already exported or imported.

For shared knowledge collections, CALC prepares uploaded content so authorized users can search it. Excerpts may be sent to an AI provider for an answer. Team permissions control who can use content within CALC. Whether and how that content may be processed externally still needs to be assessed.

For licenses, quotas, invoices, and support, we process your account, company, contract, usage, and billing details. On our servers, we link your verified CALC account to a contact in our customer management system, NetGenium, or create a contact there. This lets us manage our business relationship with you and use consistent customer details across these activities.

We use Stripe for online payments where offered. Depending on the payment method, Stripe processes the necessary customer, order, invoice, and payment data. We use Mailgun to send emails such as account confirmations, password recovery messages, and sharing messages you request. This involves processing the recipient's address, the necessary message content, and delivery details.

Where applicable, we rely on Article 6(1)(b) GDPR for contract performance, Article 6(1)(c) for legal obligations, and Article 6(1)(f) for our legitimate interests in consistent customer management and handling general inquiries. Advertising serves a different purpose and needs an appropriate legal basis. Accepting cookies does not sign you up for a newsletter.

Chat history, uploaded files, memories, and analysis results follow different deletion rules. So do account and billing records, consent records, technical logs, and backups, because we need them for different purposes. Deleting a visible chat therefore does not automatically delete all related data at the same time.

Under our existing deletion policy, chat-related analysis results are deleted with the corresponding chat. You can delete saved memories separately. Business records that we must retain by law may remain after an account is deleted. Their use must then be limited to the necessary purpose.

Backups and data stored by external services need to be considered separately. For questions about storing or deleting your data, contact [email protected].

Protecting personal data requires restricted access, secure sign-in, protected transmission, and clear procedures for handling incidents. The technical and organizational measures needed depend on the risks of the processing. When we send data to an AI service we engage, another recipient receives that data. This does not publish it on the internet.

Submit only content you need for your task and are authorized to process. Follow confidentiality agreements, employer and client requirements, and any special secrecy obligations. Data does not automatically become anonymous when you remove a name. Special categories of personal data, such as health information, require additional legal conditions. Dlubal remains responsible for its own data protection duties regardless of these instructions.

We may also receive personal data through your sign-in provider, existing Dlubal customer relationships, or content other users enter in CALC. Examples include contact and company details or data in project documents.

Where we are responsible for this data as controller, the information requirements in Article 14 GDPR may also apply. If we must notify someone individually, publishing this notice does not automatically replace that notification.

We update this notice when our data processing or the legal requirements change. The date at the top shows the current version. Before we continue processing data for a materially different purpose, we explain the change. Where needed, we ask for new consent. Updating this notice alone does not replace that consent.

You can read more about the legal requirements in the General Data Protection Regulation, Section 25 TDDDG on device protection, and the European Commission guidance on standard contractual clauses. These sources explain general requirements. We must also provide specific information about how CALC processes data.

If you consent to usage analytics, we can record that you opened the invitation dialog from a CALC reminder email after signing in. Within 30 days, we associate an invitation created from that dialog with the email. We store the campaign email identifier, dialog opening time, and invitation association on our servers. We use the sending status and a confirmed registration to assess campaign results. This association does not prove that the email caused the registration.

You can still use invitations without analytics consent. After you withdraw consent in the cookie settings, we do not record further dialog openings or new invitation associations. Product news consent remains separate. Measurement data is visible only to authorized administrators and is stored with the campaign history.

CALC uses cookies for your sign-in session, your cookie consent choice, and an anonymous session ID. We store your consent choice for 12 months. A random device token is stored for up to 400 days and is used only to enforce the free quota.

These cookies come from us, contain no tracking identifiers, and are not shared with third parties. The legal bases are Section 25(2) TDDDG and Article 6(1)(b) and (f) GDPR.

To keep the free plan available, we limit the free quota per device rather than per email address. We primarily use the random device token described above.

We also calculate a device fingerprint, but only when specific risk signals appear. Examples include a nearly exhausted free quota, an account switch on a known device, or a device token deleted after prior use. The raw fingerprint never leaves this processing step. We store only a pseudonym generated with a key (HMAC), delete it after 180 days of inactivity, and never use it for analytics, advertising, or marketing.

A match never blocks you automatically. It can only trigger an additional verification step. The legal basis is our legitimate interest in protecting the free plan under Article 6(1)(f) GDPR. We have documented the necessity assessment under Section 25(2) TDDDG internally.

With your consent, we load Microsoft Clarity to understand how people use the app. Clarity records session replays and creates heatmaps showing which areas people use most. All text and media content is masked in recordings, including chats, project and file names, and profile data. We do not send Clarity your email address, account ID, or device identifiers. Advertising storage is permanently disabled.

The recipient is Microsoft Ireland Operations Ltd. Transfers to the US are safeguarded by the EU standard contractual clauses and Microsoft's EU Data Boundary commitments. Microsoft retains recordings for up to 30 days and aggregated usage data for up to 13 months.

The legal basis is your consent under Article 6(1)(a) GDPR. You can withdraw it at any time in the cookie settings. CALC then stops Clarity processing and reloads. We cannot technically delete cookies that Clarity sets under the clarity.ms domain, but the withdrawal signal stops their use.

CALC reports technical errors to our own servers so we can find and fix faults. A report contains the error message, technical stack trace, affected page, and program version. The stack trace shows the program steps that led to the error. These reports are sent in every session regardless of consent because they are necessary for secure and reliable operation.

CALC does not read or set cookies for this, and it does not send a session or user identifier. We do not link the report to your account. Your IP address is processed when the connection is established. Email addresses are automatically removed from error messages before storage.

The data stays on our own servers, with no other recipients. The legal basis is our legitimate interest in secure and reliable operation under Article 6(1)(f) GDPR.

With your consent, we also load Grafana Faro. It adds loading times, browser and device details, and a temporary session ID to the error monitoring described above. This helps us understand which steps led to a fault. It also records performance traces for requests to our own servers (timing and URL, without query parameters), so we can link a slow page to the server processing behind it. Traces are deleted after 7 days.

The session ID exists only in the open tab's memory and is not saved on your device. We do not send your email address, account ID, or device identifiers. Faro sends data to our own servers; no third party is involved.

The legal basis is your consent under Article 6(1)(a) GDPR. You can withdraw it at any time in the cookie settings. CALC then stops collection and reloads.

After you sign in with a verified email address, we link your CALC account to a contact in our customer relationship management system (CRM), or create a contact there. This allows licensing, support, and billing to use the same customer details.

The matching takes place exclusively between our servers. The email address used is never included in identifiers for usage analytics or abuse protection. Any marketing would be a separate processing activity with its own legal basis and an opt-out.

The legal bases for account management are Article 6(1)(b) GDPR (contract) and Article 6(1)(f) GDPR (our legitimate interest in consistent customer management).

We have completed chats analyzed automatically by an AI model to identify technically inadequate answers and difficulties using CALC. The model reviews message text, your ratings and feedback, and which tools were used. Uploaded files, images, and drawings are excluded from this analysis.

The results contain verbatim quotes from your messages. Only Dlubal Software administrators can view them. We do not share the results with third parties or our CRM. The analysis is currently performed by OpenAI, one of the contracted AI providers that also process chats in CALC (Google, OpenAI, Anthropic). The analysis serves a separate purpose. We delete it with the corresponding chat. There is no separate retention period.

Mood analysis is switched off by default and also serves a separate purpose. It examines how your mood changes during a chat and whether frustration concerns the product or the task. It runs only if you turn it on under Settings → Personalization. Turning it off stops the analysis and deletes the frustration signals already derived from your sessions. The same switch covers a sales analysis that would identify needs and interests for our sales team. That analysis has its own legal basis and is currently inactive.

Product analysis relies on our legitimate interest in improving CALC and its answers (Article 6(1)(f) GDPR). Mood analysis requires your consent under Art. 6(1)(a) GDPR, which you can withdraw at any time.

You can object at any time to product analysis under Settings → Personalization. We then stop future analysis. The analyses already created for your chats, including saved quotes, are deleted.

Alongside the AI provider for your chats, we use TypeSafe's decision model Jev through OpenRouter. Jev assigns your messages to the matching skill. For the purposes described in the previous section, it also evaluates questions and their answers. For example, it checks whether an answer addresses the question, whether it gives design values without advising you to have them checked, or whether you want to speak with a person.

If you turn on Memories under Settings → Personalization, Jev also checks whether you mention a lasting preference, such as a design code or response style. We compare that classification with the memories saved by the AI assistant. Jev does not create memories itself. Turning memories off stops this check.

For this, we send the text of your message, the answer, and the previous answer to OpenRouter and TypeSafe. We remove API keys first. For the analysis, we store only the classifications, not the text. Only Dlubal Software administrators can view the classifications. We delete them with the chat, or at the latest 90 days after the relevant message.

You can object to this evaluation using the same switch under Settings → Personalization. We then stop the evaluation and delete the classifications already saved for your chats. Jev does not evaluate mood. The legal basis is our legitimate interest in improving CALC and its answers (Article 6(1)(f) GDPR).

While the voice agent is on, it continuously sends your microphone audio to Google's Gemini Live API. This includes time when you are not speaking. The audio passes through a Dlubal relay server, which forwards it without storing it.

Google receives the audio, your selected language, and the text of questions and answers from your conversation with CALC. Google returns synthesized speech and a written transcript. Dlubal does not store the audio. The transcript and session cost are saved in the chat. We keep a technical conversation log for 24 hours to investigate errors.

The legal basis is your consent under Article 6(1)(a) GDPR. We ask for it when you first turn on the voice agent and record it with a timestamp. You can withdraw it here at any time. After withdrawal, no new session can start. A session already running ends at its next reconnection. Withdrawal does not affect the lawfulness of earlier processing.

We use Cloudflare Turnstile in the registration form to block automated sign-ups. Turnstile checks technical browser signals. We receive only a token indicating whether the check passed.

The recipient is Cloudflare, Inc., under EU standard contractual clauses. The legal basis is our legitimate interest in preventing automated sign-ups under Article 6(1)(f) GDPR.

Subject to the legal conditions, you can ask to access your personal data (Article 15 GDPR), have it corrected (Article 16), have it deleted (Article 17), or restrict its processing (Article 18). You also have a right to data portability (Article 20). This concerns data you have provided that we process automatically based on your consent or a contract.

Object to processing: If we process data based on legitimate interests, you can object at any time for reasons relating to your particular situation (Article 21 GDPR). We then stop processing unless we can show compelling legitimate grounds that override your interests, rights, and freedoms, or we need the processing for legal claims. You can object to direct marketing and related profiling at any time without giving those reasons.

Withdraw consent: You can withdraw your consent at any time for the future. This does not affect the lawfulness of earlier processing. For Clarity and Faro, use the cookie settings. For product analysis and personalization, use the account settings described above. The voice agent has a separate consent control. You can also contact our data protection officer.

We normally handle requests within one month. If the law allows us to extend that period, we tell you within the first month and explain why. If we have reasonable doubts about your identity, we may ask for the additional details needed to verify it.

You can complain to a data protection supervisory authority, particularly where you usually live or work, or where you believe an infringement occurred. The authority responsible for Dlubal is the Bavarian State Office for Data Protection Supervision (BayLDA).

If you connect an AI assistant such as Claude to your CALC account, that assistant sends questions to CALC on your behalf and receives Mia's answers. The provider of the assistant you chose (for example, Anthropic for Claude) therefore receives the answers and whatever you include in your questions, under its own terms. CALC stores these chats in your account like any other chat and charges credits for each answer.

We store which apps you connected, when you connected them, when they were last used, and the access tokens in hashed form only. You can disconnect an app at any time under Settings, API Keys; its access ends immediately.

The legal basis is Article 6(1)(b) GDPR (contract), because you set up the connection yourself.

On plans from Starter up, read aloud sends the text of the answer to Google's Gemini API, which returns synthetic speech. The answer can contain details of your project. Dlubal stores the generated audio for up to 30 days so that playing it again costs nothing, and deletes it as soon as you delete the chat. Legal basis is the performance of the contract (Art. 6(1)(b) GDPR). On the Free plan, and whenever Mia's voice is unavailable, your browser's own speech output reads the answer instead; depending on the browser, its online voices may process the text at the browser vendor (for example Google or Microsoft).

When you publish your own skill in CALC or change a published skill, we classify and review it automatically. This lets us organize the skill library, spot duplicate skills and improve their quality. Dlubal's own skills go through the same review.

TypeSafe's decision model Jev, accessed through OpenRouter, assigns the skill to categories such as task, domain and the closest Dlubal skill. An AI model from Anthropic writes a description, strengths, weaknesses and suggestions for improvement. For this, we send the skill's name, description and instruction text. We remove API keys first. Your chats and files are not affected.

We store only the classification and the review result. Only Dlubal Software administrators can view the results. We delete them when you delete the skill. We adopt content from your skill into Dlubal's skills only after consulting you.

The legal basis is our legitimate interest in improving CALC and its skills (Article 6(1)(f) GDPR). You can object to this review at any time using the contact details above. We will then exclude your skills and delete the results already created.