This page is a public summary of data protection considerations for Risphai. It explains processing purposes, risk themes, and safeguards without publishing confidential technical details. It is not a complete statutory Data Protection Impact Assessment (DPIA), a certification, or a finding that no further assessment or consultation is required.

The Privacy Policy provides the detailed notice about data categories, recipients, retention, and rights. The Terms of Service govern use of the Service. Mandatory data protection obligations remain applicable.

Risphai is an AI assistant available through its website and mobile apps. It supports text responses, research, code and file creation, media generation, voice features, optional personalization, projects, and scheduled tasks.

Requests are processed by available AI services, including hosted open-source models, according to the feature and selected settings. Outputs are generated by AI and may be inaccurate, biased, or unsuitable. Users should review them before relying on or sharing them.

• Accounts: Authentication details, account settings, and optional profile information needed for access and personalization.
• Content: Messages, relevant conversation context, uploaded files, extracted content, generated results, saved prompts, memories, projects, and scheduled task details.
• Audio: Live audio for voice processing and uploaded audio for transcription. Saved transcripts and generated audio follow the applicable conversation or file retention rules.
• Payments: Purchase and subscription identifiers, entitlement status, credit usage, and financial records. Risphai does not store full card numbers or CVVs.
• Technical and safety data: Request and device information, consent preferences, limited security and enforcement records, and reliability diagnostics.
• Native notifications and diagnostics: Push registration information and notification content when enabled, plus mobile crash information and installation identifiers when optional crash diagnostics are enabled.

Do not provide unnecessary sensitive information or third-party data without authority. The Service is not designed for regulated handling of sensitive or criminal-offence data.

Risphai processes account and content data to supply the requested feature. Relevant content is disclosed to AI or retrieval providers; saved account data and content are handled by database and file-storage providers. Payments, email, optional sign-in, embeds, native notifications, and diagnostics involve their respective recipients.

Incognito limits saved conversation history but still involves temporary request and output processing, session recovery, and cleanup. Providers can retain data separately. Privacy Mode restricts eligible text-model inference; image generation and editing, video and audio generation, search, retrieval, voice, and required safety checks follow separate retention rules. OpenRouter requests enforce allowed inference providers and, where applicable, its per-request ZDR filter. fal requests opt out of payload storage, set output files to expire after one hour, and request deletion of video and audio request data once the result is stored by Risphai.

Provider identities, purposes, data categories, and processing locations are listed in the Privacy Policy. No specific model routing or internal implementation is promised by this summary.

Where the GDPR applies, processing relies on contract performance for requested services, legal obligations for required records, consent where needed for optional processing, and legitimate interests for proportionate security, abuse prevention, and reliability measures, subject to users' rights. Swiss law also requires lawful, transparent, proportionate processing.

Acceptance of the Terms or this notice does not itself establish consent for all processing. Consent may be withdrawn where it is the basis. Sensitive-data processing requires its own applicable legal condition; entering information is not a universal substitute for that condition.

• Unauthorized access or disclosure: Account and file access controls, authentication, encrypted network connections, and restricted operational access help reduce exposure.
• Provider and transfer risks: Data can be handled by external recipients and in other countries. Applicable processing agreements and lawful transfer mechanisms are needed; a privacy policy alone is not a safeguard.
• Over-retention and data loss: Saved and temporary content have different lifecycles. Deletion, cleanup, retained financial records, and backup handling must reflect their purposes and legal obligations.
• AI and moderation errors: Automated outputs and safety classifications can be wrong. Users should verify outputs and may challenge restrictions through support and request review by a person.
• Sensitive content and sharing: User submissions, share links, and notification previews can expose personal information. Submit only necessary content and review it before sharing.

These are general risk themes, not a completed residual-risk assessment. Controls reduce risk but do not eliminate it. A statutory DPIA, where required, must assess the actual processing, necessity, proportionality, affected people, risks, and safeguards before that processing begins.

Optional personalization can be edited, disabled, or removed. Disabling memories stops their use in new requests without deleting existing entries. Relevant context is supplied to providers as needed for the feature; avoid including unnecessary personal information.

Saved account content generally remains until deleted. Temporary incognito request data is scheduled for cleanup after completion or abandonment, as described in the Privacy Policy. Payment records, enforcement records, backups, and provider-held records have separate retention rules. Removing an account link from a financial record does not necessarily make it anonymous.

Software bugs, outages, failed cleanup, or provider delays may cause unintended storage or retention beyond normal periods, including with incognito or Zero Data Retention. Identified failures require investigation, restricted unnecessary use, and deletion as soon as reasonably practicable. These limitations do not permit indefinite retention or remove statutory deadlines and data protection duties.

See Data Retention for periods, triggers, and exceptions.

Account controls provide data export, deletion, profile editing, content management, privacy settings, and optional cookie preferences. You may also request access, correction, erasure, restriction, portability, or objection where applicable law provides those rights.

For requests or to challenge an automated restriction, contact support@risphai.com. You may request review by a person. We may verify identity proportionately and respond within statutory deadlines, explaining any lawful extension. You can complain to the applicable supervisory authority without our permission.

If you are located in the EU/EEA, you may also contact our appointed GDPR Article 27 representative, Complico Consulting GmbH. Its address, email and secure DSAR portal are provided in the EU/EEA GDPR Representative section of our Privacy Policy.

Safeguards include encrypted network connections, account authentication, optional multi-factor authentication, access restrictions, content safety screening, and limited diagnostic reporting. Operational access and deliberately shared content follow the relevant permissions.

No system can promise absolute security or perfect moderation. Failures do not remove our obligations to investigate incidents, protect personal data, and provide legally required notifications. Confidential security details may be withheld only consistently with applicable disclosure duties.

Recipients include hosting and storage services, authentication and email providers, AI and hosted open-source model services, search and page-reading services, payment and subscription services, native notification and crash-reporting services, and optional embedded-content providers.

Some act as processors under our instructions; others also act as independent controllers for their own services. The Privacy Policy identifies the providers and applicable data categories. International transfers require safeguards appropriate to the recipient and destination.

OpenRouter supplies AI routing and fallback inference under our data processing agreement. Relevant content and an internal user identifier are sent through OpenRouter to selected inference providers. Depending on the model, these can include OpenAI, Microsoft Azure, Google AI Studio, Google Vertex AI, Anthropic, Amazon Bedrock, xAI, InferenceNet, Morph, DeepInfra, CoreWeave, DigitalOcean, NextBit, Novita, Ionstream, Makora, Relace, Wafer, Baseten, Together, Modal, Parasail, Venice, Near AI, Phala, SiliconFlow, Reka, Sail Research, and Prime Intellect. OpenRouter retains request and usage metadata, while content logging and product-improvement use must remain disabled in our account. ZDR applies to eligible inference endpoints and does not remove metadata retention or the need to assess subprocessors and international transfers.

fal supplies media generation under our data processing agreement, including MiniMax video models and fallback processing for supported Google, xAI and ElevenLabs image, video and sound-effect models. Media and audio prompts, source images and generation settings are sent to fal, which may pass the content needed for processing to the model developer, including Google, xAI, MiniMax and ElevenLabs. Requests opt out of fal payload storage and output files are set to expire; media and audio generation are outside ZDR. The model developers extend the recipient and transfer chain and need assessment under the applicable agreements.

Processing risks and disclosures need review when features, recipients, data categories, security conditions, or applicable law change materially, and following relevant incidents. This public summary does not certify completion of an internal assessment or establish a future assessment date.

Publishing a summary does not replace an internal assessment, processing records, provider agreements, lawful transfer arrangements, or any required supervisory consultation. Whether a statutory DPIA is necessary depends on the actual nature, scale, context, and risk of processing.

Further information can be requested through support. Required disclosures and supervisory access remain available subject to applicable law and lawful protection of confidential information.