app.klang.ai — 62/100 (Risque moyen pour la vie privée)
Dernière analyse
Le contenu du rapport (résumé, constats, citations) a été généré en anglais et n'est pas localisé.
Klang AI AB · klang.ai
Détails du rapport
Risque moyen pour la vie privéeKlang AI collects potentially sensitive audio and video content, shares it with multiple AI providers, and reserves the right to use your uploaded data to improve its own algorithms without offering an opt-out — though it does keep most processing in the EU and enforces zero data retention with its AI sub-processors.
Klang AI AB, a Swedish AI transcription and meeting-recording service, processes a broad range of personal data including potentially sensitive content from audio/video files. The policy is commendably transparent about its sub-processors and has zero data retention agreements with LLM and Speech AI providers. However, it claims the right to use uploaded content to improve its own algorithms without a clear opt-out, collects credit card numbers directly (despite using Stripe), and transfers data to at least one US provider (ElevenLabs) without describing transfer safeguards. Retention is capped at three months post-account termination, which is relatively strong.
Évaluation par catégorie
Répartition de la politique selon les principaux domaines de conformité. Bon = solide, moyen = mitigé, faible = préoccupant.
Collects credit/debit card numbers directly despite using Stripe as a payment provider, and broadly defines collectible data to include any personal information in uploaded audio, video, or text files.
Clearly lists all sub-processors by name, purpose, and location; explicitly states zero data retention agreements with AI providers; structured policy with a useful summary.
Extensive sub-processor list is disclosed with detail, but data flows to at least one US provider (ElevenLabs) without transfer safeguards, and integration partners like Slack, Dropbox, HubSpot are also US-based.
ElevenLabs (USA) receives audio data for speech AI, and multiple integration partners are US-based, yet the policy is completely silent on SCCs, adequacy decisions, or any other Chapter V safeguard.
Third-party LLM and Speech AI providers are contractually barred from training on or retaining user data, but Klang itself reserves the right to use uploaded content to improve its own algorithms with no opt-out mechanism.
GDPR rights (access, rectification, erasure, restriction, portability, objection) are listed with supervisory authority links, but there is no automated self-service mechanism — users must email support@klang.ai.
Constats clés
Clauses notables, problèmes ou bonnes pratiques identifiées (critiques en premier)
Algorithm improvement on user data without opt-out
Section 2 states: 'Uploaded content which may include personal information can be used to improve our services and algorithm.' This means Klang can use potentially sensitive audio/video content to train or refine its own models. No opt-out is offered, and the legal basis for this processing is unclear — it is not separately listed in Section 3's legal bases. For an AI transcription service handling meeting recordings, this is a significant concern.
US data transfer without documented safeguards
ElevenLabs, Inc. is listed as a Speech AI provider based in the USA. Audio data is shared with ElevenLabs for transcription and speaker identification. The policy does not mention Standard Contractual Clauses, an EU-US Data Privacy Framework certification, or any other GDPR Chapter V transfer mechanism. This is a compliance gap that could expose Klang to enforcement risk.
Credit/debit card numbers collected directly
Section 1 lists 'debit/credit card numbers' as personal information collected. Since Klang uses Stripe as its payment provider, full card numbers should never reach Klang's servers. Collecting and storing card numbers would create unnecessary security risk and likely violates PCI DSS and the data minimization principle.
Sensitive data processing consent mechanism is vague
Section 1 acknowledges processing 'potentially sensitive information contained within text or files (such as audio and video)' and says this is done 'with your consent or as otherwise permitted by applicable law.' However, the policy does not describe how consent is obtained — whether at upload, at account creation, or via a separate mechanism. For a service whose core function involves processing audio/video of meetings, this lack of specificity is concerning.
Zero data retention agreements with AI sub-processors are a strong positive
The policy explicitly states that LLM providers (Scaleway, Mistral, Microsoft Ireland, AWS Europe, Berget AI) and Speech AI providers (ElevenLabs, Pyannote) operate under zero data retention agreements where 'Data is not used for training and is not retained after processing' and 'Audio data is not used for training and is deleted immediately after processing.' This is above-average transparency for an AI service.
Short retention cap is notable
Section 6 states: 'No purpose in this notice will require us keeping your personal information for longer than three (3) months past the termination of the user's account.' This is a relatively short and clearly defined retention period, which is positive for GDPR compliance.
Synthèse pour l'utilisateur
Your meeting recordings and transcriptions could be used to train Klang's own algorithms, and some of your data flows to a US-based speech AI company with no explained legal safeguard — but third-party AI providers contractually cannot retain or train on your data.
Posture de conformité
Klang AI demonstrates partial GDPR compliance: it identifies legal bases, lists sub-processors with specificity, and caps retention. However, it has gaps around international transfer safeguards, lacks a clear opt-out for its own algorithm improvement processing, and collects payment card data that it likely should not be storing directly. No DPA reference or DPIA mention is made.
Transferts UE
Most processing stays within the EU/EEA (Scaleway in France, Mistral in France, Microsoft Ireland, AWS Europe, Berget AI in Sweden, Pyannote in France). However, ElevenLabs Inc. is US-based, and the policy is silent on transfer mechanisms such as Standard Contractual Clauses or an adequacy decision. This is a compliance gap under GDPR Chapter V.
Signaux détectés
Données et pratiques spécifiques identifiées dans le texte
Extraits probants
Citations directes de la politique à l'appui de ces constats
Uploaded content which may include personal information can be used to improve our services and algorithm.
LLM Model providers, we work with the following AI providers who have strong data protection policies and zero data retention agreements. Your data can be shared with these to perform LLM requests. Data is not used for training and is not retained after processing.
No purpose in this notice will require us keeping your personal information for longer than three (3) months past the termination of the user's account.
ElevenLabs, Inc., USA
debit/credit card numbers
When necessary, with your consent or as otherwise permitted by applicable law, we process the following categories of sensitive information: any potentially sensitive information contained within text or files (such as audio and video) that are submitted to the service.
Manquant ou flou
- No description of international transfer safeguards (SCCs, DPF, BCRs) for US-based sub-processors
- No opt-out mechanism for Klang's own algorithm improvement using uploaded content
- No explanation of how consent for sensitive data processing is obtained in practice
- No reference to a Data Processing Agreement (DPA) for B2B customers
- No mention of Data Protection Impact Assessment (DPIA) despite processing sensitive audio/video at scale
- No automated self-service portal for exercising GDPR rights
- No details on cookie categories or tracking technologies (referenced to separate Cookie Notice)
Questions à poser
- What specific transfer mechanism (SCCs, EU-US DPF certification, etc.) do you rely on when sharing audio data with ElevenLabs in the United States?
- How exactly is consent for processing sensitive audio/video content obtained — is it at upload time, in the sign-up flow, or elsewhere?
- Can users opt out of having their uploaded content used to improve Klang's own algorithms, and if not, what is the legal basis for this processing under GDPR?
- Do you actually collect and store full debit/credit card numbers on your own servers, or does Stripe handle all card data via tokenization?
- What does 'improve our services and algorithm' entail — does this involve training machine learning models on user-uploaded audio/video content?
- Is a Data Protection Impact Assessment available for the processing of sensitive audio/video content at scale?
Partager cette analyse
Toute personne disposant de ce lien peut consulter le résultat ci-dessus.
Conçu par DentroChat
Chat IA 100 % européen pour tous
Discutez avec l'IA, travaillez avec des fichiers, générez des images et cherchez sur le web. Les données restent en Europe.
Autres politiques de confidentialité analysées
Tout affichercursor.com
62/100
Risque moyen pour la vie privéeCursor (Anysphere) promises not to use your code inputs for AI training by default and doesn't sell your data, but it collects a sweeping range of personal and usage data, ships it to the US with vague transfer safeguards, and leaves key details like retention periods and legal bases unspecified.
Voir le rapport →eustella.com
62/100
Risque moyen pour la vie privéeeustella fait de fortes promesses en matière de confidentialité — aucune vente de données, aucun partage avec des tiers, aucun entraînement d'IA sur vos données, et tout le traitement reste dans l'UE — mais il s'agit d'un texte marketing, et non d'une politique de confidentialité contraignante, et des détails cruciaux sur la collecte des données, la conservation, les sous-traitants et les droits des utilisateurs sont absents du texte fourni.
Voir le rapport →lhv.com
62/100
Risque moyen pour la vie privéeLHV Bank collecte une vaste gamme de données personnelles et financières, les partage largement avec des agences de prévention de la fraude et des intermédiaires de paiement, et s'appuie fortement sur l'intérêt légitime pour le marketing et l'analyse — mais elle couvre les droits du RGPD et utilise un examen manuel pour les décisions de profilage.
Voir le rapport →cake.com
62/100
Risque moyen pour la vie privéeCAKE.com fournit des droits sur les données standard pour l'UE et des garanties de transfert, mais collecte des données de surveillance sur le lieu de travail très intrusives — comme des captures d'écran, la localisation en arrière-plan et l'utilisation des applications — au nom des employeurs, qui agissent en tant que responsables du traitement pour leurs employés.
Voir le rapport →