Security and privacy
Your data stays yours.
A text is like a letter. You want to know who carries it, who opens it and where it lies afterwards. So I show you the whole route: your text goes into your own workspace at RhetorIQ, from there to the AI, then back to you as a draft. After that I read it and refine it by hand. Every stage is secured by contract and by technical means, and where I cannot promise something, I tell you openly.
Request a callThe three security levels
- Your textBrief and keywords from your workspace
- RhetorIQServer and database at Render, location Frankfurt (EU)
- AI providerAnthropic (USA) as data processor
- Back to youThe draft appears in your workspace
- Lorena readsand refines the text by hand
Gold: processed in the EUSilver: processed in the USA
What really happens
Four doors, and a bolt for each.
The AI
For text generation I use the interface from Anthropic. According to Anthropic's documentation, content sent through this interface is never used for training without explicit permission. Zero Data Retention, meaning no storage at all at Anthropic, is available to organisations on request.
The limit belongs here too: content that Anthropic flags as abusive can be kept for up to two years.
The platform
- Every client has an own workspace. The data is separated.
- Team members have roles: viewers read, editors write, admins manage.
- Deleted client workspaces sit in a recycle bin for 30 days and are then deleted completely.
The retention periods
- Sent emails lose their content and attachments after 7 days, the entries disappear after 90 days.
- The AI error log is deleted after 90 days.
- The periods run automatically every day.
Protection in the engine room
- Access keys of third parties, for example for calendars or newsletter accounts, are stored encrypted in the database (AES-256).
- Texts from outside sources such as reference documents are marked as material for the AI. It is told to ignore instructions inside them.
- Keys and credentials are removed from logs and error messages.
The human in the process
I read every text and refine it by hand. That is your quality advantage and also access to your content. Both belong in the same line.
Contracts and transparency
In black and white, who processes what.
A data processing agreement under the revised Swiss Data Protection Act and the full list of sub-processors are available on request. These providers are in use today:
- RenderHosting and database, location Frankfurt (EU)
- AnthropicAI processing of your briefs (USA)
- BrevoSending the platform's emails (EU, France)
- AssemblyAISpeech recognition for uploaded audio and video files (USA)
- StripePayments and subscriptions
- SentryMonitoring of technical errors, without request content and cookies
Three security levels
You choose the lock that fits your door.
Each level says clearly what is built today and what only comes into being on request. Only what is agreed by contract or already in operation is promised.
Level 1
Standard
Available- Data processing agreement and open list of sub-processors
- Transparency on data route, retention periods and access
- Training for your team
- Traffic-light rules for daily work (see below)
Level 2
Protected
- Data guard: recognises sensitive details such as AHV number, IBAN or email address and replaces them with placeholders before text goes to the AIIn progress
- Zero Data Retention at AnthropicOn request
- Claude in Europe through a European cloud providerOn request
Level 3
Sovereign
On requestYour own instance with hosting in Switzerland and a language model operated in Switzerland.
To be honest: such a model is smaller. The text quality can therefore be lower than with Claude. We test this beforehand with your own texts.
Traffic-light model for staff
Three colours everyone understands at the coffee machine.
Before every input, your team asks one question: what colour is this content?
Green
Use freely
Public content or texts meant for publication.
Examples: press release, website text, LinkedIn post.
Yellow
Only in the platform
Internal information without personal data. It belongs in the platform and in no other AI tool.
Examples: draft of an internal memo, talking points, strategy paper without names.
Red
Never into the AI
Personal data, health data, AHV numbers, contracts, content under professional secrecy and passwords.
Examples: patient file, employment contract, login details.
Questions
Briefly and honestly answered.
What happens to my texts?+
Your text sits in your own workspace. To create the draft it is sent to Anthropic, the draft comes back, and I read it and refine it by hand. Stored texts stay in your workspace until you have them deleted.
Is the AI trained on my texts?+
According to Anthropic's documentation, content sent through the interface is never used for training without explicit permission. What RhetorIQ learns from your corrections applies only to your workspace.
Where is my data stored?+
RhetorIQ's server and database are at Render. Server and database are in Frankfurt am Main (EU). Render is a company based in the USA. To create a draft, the text goes to Anthropic in the USA. The platform's emails are sent by Brevo from the EU. You receive the full list on request.
Who sees my texts?+
You and your team, depending on role. Also I as your consultant, because I read every text. Anthropic processes the texts on your behalf. Other clients never see your data.
What about the US CLOUD Act?+
Anthropic is a US company. The CLOUD Act allows US authorities, under certain conditions, to demand data from such companies even when it is stored outside the USA. For data under professional secrecy I recommend Level 3 and your own risk assessment, ideally with your legal counsel.
Can I have my data deleted?+
Yes. On your request I delete your workspace completely, after an export of your data if you wish. I answer the request within 30 days. Deleted data can remain for a short time in the hosting provider's backups until they are overwritten.
Do you offer staff training?+
Yes. Training is part of Level 1. Your team gets to know the traffic-light rules for daily work.
Call
Twenty minutes, then you know the route of your data.
We talk through your situation and the fitting security level. The call is free and without obligation. I provide the data processing agreement on request.