Privacy policy

PILOTIA OOD is a Bulgarian limited liability company (OOD), registered in the Bulgarian commercial register under EIK 207520710. Fully paid-up share capital: 200 000 EUR.

Registered office: 93 Ekzarh Yosif Street, Oborishte district, 1000 Sofia, Bulgaria. Replybot is provided and operated by PILOTIA OOD.

Contact for legal and personal data enquiries: contact@replybot.org; contact form: www.replybot.org/en/contact. You may also write to the registered office.

PILOTIA OOD is the data controller for data needed to operate the website, accounts, subscriptions, security, and its relationship with business customers.

For data relating to people who interact with a voice agent deployed by a business customer, by phone or in the browser, that customer determines the purposes of processing and acts as data controller. PILOTIA OOD processes the data on its behalf as a processor to provide the service.

For any question regarding the processing of your data, you can contact us through the dedicated form on the platform.

As part of using our service, we collect the following categories of data:

• Identification data: first name, last name, email address, account ID.
• Connection data: IP address, date and time of connection, browser type.
• Usage data: telephone numbers linked to agents, call statistics and usage, instructions, knowledge base, opening hours and voice-agent settings.
• Payment data: handled exclusively by our payment provider Polar (no banking data is stored).
• Voice-agent call data: caller number, display name where available, date and time, duration, status, call audio recording, transcript, summary, actions performed by the agent, and any transfer number.
• Google Calendar integration data, when enabled: Google email address, selected calendar identifier, permissions and configuration, OAuth refresh token encrypted at rest with AES-256-GCM, and appointment information needed for requested actions.
• Pipecat technical data: providers, models and voices used, participating agents and handoffs, STT/LLM/TTS usage (audio duration, characters or tokens depending on the provider) and call cost estimates.

Your data is processed for the following purposes:

• Provide and maintain the service: account creation, voice-agent configuration, call processing and AI voice response generation.
• Answer calls, hold a real-time conversation, perform actions configured by the business customer, provide the call history, and support follow-up from the dashboard.
• Handle billing and subscription tracking.
• Improve the service and detect technical issues.
• Diagnose and follow up on incidents affecting calls, voice agents or connected integrations.
• Comply with our legal and regulatory obligations.

For customer accounts, processing is mainly based on performance of the contract with PILOTIA OOD, our legal obligations, and our legitimate interest in securing and improving the service. Non-essential cookies, if used, rely on consent.

For end contacts' conversations and calls, the business customer selects and documents the legal basis appropriate to its purpose, such as performance of a contract, a legal obligation, legitimate interests following a balancing test, or consent where required. PILOTIA OOD does not record calls for its own marketing purposes.

When someone calls a number on which a voice agent is enabled, the caller's voice and the agent's voice are processed in real time to conduct the conversation. Calls are recorded by default. At the end of a call, an MP3 audio file, a text transcript, and a summary are stored in the business customer's workspace, together with related metadata and agent actions.

This content may reveal personal information and may include sensitive data if the caller chooses to disclose it. Callers should not provide payment-card data, passwords, or other secrets that are not necessary for their request.

The business customer must inform callers at the start of the call that they are interacting with AI and that the call is recorded and transcribed. It must state the purpose, legal basis, retention period, recipients, applicable rights, and a way to access the full notice. Where the law requires it, the customer must also obtain consent or offer a non-recorded alternative.

Browser voice tests require microphone permission. Audio is transmitted to the voice service during the session; a test may also produce a recording, transcript, summary and action history linked to the tested agent. A session identifier may replace the phone number. Call privacy and retention rules also apply to these tests.

Depending on the selected engine, conversations are processed by a real-time voice service or separate services connected through Pipecat. Pipecat is a software library running on our backend: using it does not, by itself, send conversations to a service hosted by the library’s publisher.

The STT service receives the caller’s audio stream and produces text. The LLM receives the transcript, agent instructions, conversation context and relevant information from tools or the knowledge base. It generates responses and may request authorized actions. The TTS service receives the text to speak and voice settings, then produces the audio sent to the caller.

The providers actually used depend on the agent configuration and enabled functions. Data is not automatically sent to every provider listed below. Summaries and other post-call processing may also use a separate model.

A call may be handled successively by several associated Pipecat agents belonging to the same business customer. Each agent can have its own prompt, tools, models and voice. The receiving agent gets a bounded history of recent exchanges and tool results, together with a handoff summary. These may contain personal data and be sent to that agent’s LLM provider.

Returning to the main agent preserves the call identity, recording and useful information. Changing agents does not create a new call or restart retention periods. Handoffs and actions may appear in the call history. A short sound may signal the change; it does not replace full information about data processing.

The following recipients are involved depending on the services enabled. AI providers receive data needed for their STT, LLM or TTS role; other providers support hosting, telephony, integrations or service administration.

• Supabase: database hosting, authentication and private storage of call recordings.
• Vercel (United States): web application hosting.
• Twilio (United States): phone-number provisioning, routing and transport of telephone calls and, depending on configuration, SMS delivery.
• Polar (United States): payment processing and billing (Merchant of Record).
• Google LLC (United States): Google Calendar API, when you connect a calendar to a voice agent for appointment booking. Our use complies with the Google API Services User Data Policy and its Limited Use requirements.
• OVHcloud: voice backend hosting on a VPS in France, stream processing and temporary storage of call files and logs.
• OpenAI: depending on configuration, STT, LLM, TTS, real-time voice conversations, assistance drafting instructions and post-call processing.
• Google: depending on configuration, STT, LLM (including Gemini through Vertex AI), TTS or real-time voice conversations. Google Calendar is a separate, optional integration.
• Deepgram: speech-to-text (STT). The Pipecat integration uses the European endpoint and requests exclusion from the model improvement program.
• Anthropic: conversational text processing (LLM) when selected.
• ElevenLabs and Mistral AI: text-to-speech (TTS), depending on the selected voice and model.
• DeepSeek, Z.ai (GLM models) and Alibaba Cloud Model Studio / DashScope (Qwen models): LLM processing when their integration is configured and selected.
• xAI: real-time voice conversations when selected for a compatible agent.
• A custom LLM endpoint may also be configured. Its operator depends on the deployment; its identity and applicable terms must be disclosed to the relevant customer before use.

Account data needed to provide the service is retained while it is used and deleted within 30 days of account closure, except data required by law or for legal claims, which is kept separately with restricted access.

By default, measured from each call date, audio recordings are retained for 30 days, transcripts for 90 days, and summaries and associated metadata for 12 months. Action histories and logs containing conversation content follow the transcript retention period. Browser test calls, including recordings, transcripts, summaries and metadata, are retained for 7 days.

An automated daily purge removes expired content. Local recording and transcript copies are also purged. Early deletion requests can be sent to the business customer acting as controller, whom we assist. At the end of the service, data is returned or deleted under their instructions; any technical return period is limited to 30 days and does not extend existing expiry dates.

The business customer checks that these periods are necessary for the disclosed purposes. Exceptional retention requires a documented justification and a specific expiry date; an ongoing subscription alone does not justify it. Backups and copies held by providers are subject to separate deletion and expiry procedures.

Billing data is kept for 10 years in accordance with French accounting obligations.

The periods above concern data retained by PILOTIA OOD. STT, LLM and TTS providers may apply separate rules to requests, technical logs, abuse prevention or model improvement under their terms and the options of the account used. Choosing Pipecat does not imply zero retention at every provider or a blanket exclusion from training. The terms applicable to a configuration can be requested through our contact form.

Hosting the voice backend in France does not mean that all processing takes place in France or the European Economic Area (EEA). Depending on providers, endpoints, models and support services, data may be processed outside the EEA. A provider’s registered office alone does not determine the processing location.

Transfers outside the EEA must rely on an applicable GDPR mechanism, such as an adequacy decision or Standard Contractual Clauses, with a risk assessment and supplementary measures where needed. Enabling a model or selecting a technical region does not replace this assessment. Contact us for the recipients, processing locations and safeguards applicable to your configuration.

Under the GDPR, you have the following rights regarding your personal data:

For data from a call to a customer’s agent, please first contact that business, which is the data controller. PILOTIA OOD assists it with requests. For your account data or help identifying the appropriate contact, use our contact form.

• Right of access to your personal data.
• Right to rectification of inaccurate data.
• Right to erasure (right to be forgotten).
• Right to restriction of processing.
• Right to data portability.
• Right to object to processing.
• Right to withdraw your consent at any time.
• Right to lodge a complaint with a supervisory authority, in particular in the Member State of your habitual residence, place of work or the alleged infringement: CPDP / КЗЛД in Bulgaria (https://cpdp.bg/en/), or CNIL in France (https://www.cnil.fr).

The site uses cookies strictly necessary for the platform to operate (authentication, session). No third-party advertising or profiling cookies are placed without your explicit consent.

PILOTIA OOD implements appropriate technical and organizational measures to protect your data against unauthorized access, alteration, disclosure, or destruction: encryption of data in transit (TLS), encryption at rest for sensitive secrets such as third-party API access tokens (AES-256-GCM), strict access control, logging of sensitive actions, and regular backups.

Call recordings are held in private storage. Playing them from the dashboard requires authentication as a user of the relevant customer and uses a temporary signed link.

When you enable appointment booking for a voice agent, you may connect your Google Calendar. The integration uses OAuth 2.0 and is strictly optional: no Google data is collected unless you complete this connection.

The permissions ("scopes") requested at consent time are:

• https://www.googleapis.com/auth/calendar.events — create, read and update events to book or move appointments.
• https://www.googleapis.com/auth/calendar.readonly — list your calendars so you can choose which one to use.
• https://www.googleapis.com/auth/userinfo.email — display the connected account's email in your dashboard.

We store the connected Google account’s email address, encrypted refresh token, permissions, selected calendar identifier and integration settings. The calendar is queried to check availability and create, find, reschedule or cancel authorized appointments. Appointment information exchanged during a call and action results may appear in the transcript, summary or agent logs; they are then subject to call retention and access rules.

Our use of Google data complies with the Google API Services User Data Policy (https://developers.google.com/terms/api-services-user-data-policy), including its Limited Use requirements. In practice:

• We use Google data exclusively to provide and improve user-facing features (availability checks, creating, modifying and cancelling appointments by the voice agent).
• We do not transfer Google data to third parties except as strictly necessary to provide the service or as required by law.
• We do not use Google data for advertising purposes.
• We do not use Google data to train, fine-tune or improve any AI model, whether ours or a third party's.
• No human reads your Google data, except: (a) with your explicit consent, (b) for security reasons (incident investigation, abuse prevention), (c) where required by law, or (d) in aggregated and anonymized form for operational purposes.

You may withdraw the access granted to our application at any time, via either of these equivalent paths:

• from the www.replybot.org dashboard, by clicking "Disconnect" in the "Tools & integrations" tab of the relevant voice agent. The token is then revoked at Google and removed from our database.
• from your Google account, at https://myaccount.google.com/permissions, by removing the authorization granted to our application.

When someone interacts with a voice agent deployed via www.replybot.org, we process information needed for the call and its follow-up: phone number where available, audio content, transcript, summary and actions performed on behalf of the business customer.

The business customer (tenant) using www.replybot.org is the data controller for their own contacts' data. As such, they must:

• determine and document a legal basis appropriate to the purposes and obtain consent where required by applicable rules, including for any subsequent marketing use;
• provide at all times a simple, free way to withdraw consent and request deletion of personal data.
• configure a clear announcement at the start of each voice call stating that AI is being used and that the call is recorded and transcribed, and provide its own privacy notice to callers;
• limit collection to necessary information, set a retention period, and restrict recordings, transcripts, and summaries to authorized personnel;
• avoid requesting sensitive data; where such processing is necessary, identify an applicable exception under Article 9 GDPR and implement enhanced safeguards.

This policy may be updated at any time to reflect changes in our services or the legal framework. Any substantial change will be notified to you by email or through the platform.

To exercise your rights or for any question regarding this policy, you can contact us through the dedicated form on www.replybot.org.