Flipatic helps publishers create videos from their content. This policy explains how we use AI, protect client materials and share responsibility for the finished videos. It supplements the client agreement and data protection documentation.
Editorial oversight and publication
In the standard workflow, an editor reviews and approves the material at each stage, including the script, voiceover and visuals. The editor is responsible for the final content and the decision to publish.
In individually agreed workflows, videos may be created automatically through the API. Flipatic delivers the finished MP4 file to the client or their system. The client is responsible for reviewing and publishing it. Flipatic does not publish videos itself or control how files are used after delivery.
Staying true to the source
The script is based on a selected article, several articles for a recap, or text pasted by an editor. AI shortens and organises the content, adapts it to the video length, and divides it into slides, while preserving the source's meaning and adding no outside information.
URL imports are technically restricted to the domains assigned to the publisher. For pasted text, the client is responsible for its provenance. AI can make mistakes. Editors can use Check against the source to compare the script with the article. This check is optional and does not replace editorial judgement.
Visuals and voices
We use article content, files uploaded by editors, stock photo libraries, Mapbox maps, embedded posts and website screenshots. Paid stock assets are accessed through the publisher’s account. The application does not offer AI image or video creation.
Voiceovers are created using external services. We use voice libraries, custom-designed voices and voices supplied by publishers. We do not clone voices ourselves. A voice supplied by a publisher may be a cloned voice; we do not currently require separate confirmation of consent from the person whose voice was cloned.
Disclosing the use of AI
Flipatic supports an on-screen label indicating AI use. The publisher sets its wording, display duration and default visibility, and can prevent editors from changing that setting. The label is not mandatory in every configuration. We do not currently add machine-readable AI markings to MP4 files.
Processing and protecting materials
External services receive the materials needed for the selected features: article content and scripts, text for translation or voiceover creation, audio for transcription, and assets for rendering. Source comparison requires both the article and the script. Content is sent without anonymisation and may contain personal data from the source article.
| Area | Current practice |
|---|---|
| Main infrastructure | UpCloud in Warsaw, Poland. |
| External services | OpenAI, Anthropic and ElevenLabs, plus rendering through AWS Lambda: processing in the USA. |
| Model training | We do not use client content to train models. We use API services under terms that exclude such use by the providers. |
| Encryption | Data is transferred over HTTPS. Stored files are not encrypted. |
| Retention | Videos and associated assets are deleted from the main system within 60 days. Backup copies may remain for up to a further 30 days. |
| Team access | Selected client support and technical team members use individual accounts to provide support and diagnose issues. All team members are bound by confidentiality agreements. |
Rights to materials
Clients are responsible for holding the rights to materials they add to the system. They receive rights to finished videos under the client agreement, subject to the licences for the assets used. Our music library contains tracks licensed for use in online and social media videos. Publishers may also supply their own music, which is available only to the designated publication. We use client videos for our own purposes only with the publisher’s consent.
Quality checks and issue handling
Several team members test changes to the system and model versions in multiple stages, using both manual and automated tests. We sometimes ask clients to assess the results. We do not block journalistic topics such as war, violence or crime. The client agreement requires content to comply with the law. We investigate violations with the client; if the issue remains unresolved, we may suspend access and terminate the agreement.
Clients report issues through the Jira support portal. Anyone outside the client organisation can report errors, misuse or rights infringements to helpdesk@autentika.pl. A designated team handles these reports. Flipatic’s Product Owner is responsible for ensuring that this policy is followed and kept up to date. We review it every six months.
