Security
AI agent security: your files stay with you.
AI data security in three points: where your data goes, what the agent never does, how you check it.
What stays with you, what goes to the AI model.
In your environment
The agent works inside your software.
It is installed in your environment, like a staff member.
- The agent
- Your software
Goes to the AI model, encrypted connection
The text of one task, nothing more.
For a task, only the text it needs is sent to the AI model, over an encrypted connection. Your data is never used to train the models. Technical retention at the provider: up to 30 days, for security, unless longer retention is required by law.
Never leaves
Your files.
They stay stored on your machines. The agent makes no copy of them elsewhere.
Where are your files stored when an AI agent works?
Your files stay stored on your machines. The agent is installed in your environment and drives your software the way a staff member would; it makes no copy of them elsewhere. For a task, only the text it needs is sent to the AI model. The scope is agreed in writing before go-live.
The IBM and Ponemon Institute Cost of a Data Breach report puts the global average cost of a data breach at $4.99M, a 12% increase over the previous year and a record high.
HIPAA and AI: what applies to a medical practice
If you run a US medical practice, HIPAA governs the patient data an agent would read.
Check what information leaves your systems, who can access it and how actions are logged. This page makes no claim of HIPAA compliance. Details for a practice: AI for medical practices.
For medical practices, we review with you whether a BAA is required and with whom before go-live.
What the agent never does.
Limits written into its configuration, not promises.
- Give legal advice or quote fees.
- Serve a legal document: that is the enforcement officer’s legal monopoly.
- Execute an authentic deed in a notary’s place.
- Make a diagnosis or give a medical opinion.
- Sign accounts or a certificate.
- Send any external message without your approval.
- Work around two-factor authentication or any security measure.
The random check, every evening.
- At the end of the day, the agent picks a few handled files at random.
- You open them: five minutes, on a sample it did not choose.
- Every move is timestamped in a log. If a file looks wrong, you can trace the whole day.
Your questions, our answers.
- How do AI agents protect client data?
- Your files stay stored on your machines, and the agent makes no copy of them elsewhere. For a task, only the text it needs is sent to the AI model, over an encrypted connection. Your data is never used to train the models.
- Is your AI agent HIPAA compliant?
- No tool settles HIPAA on its own: it depends on how it is deployed and governed. Before go-live, we review with you whether a BAA is required. This page makes no claim of HIPAA compliance.
- Can AI make things up?
- Here it copies a document into a piece of software. Every value is checkable against the source, and you review a draft.
- Will the agent replace my staff?
- It is an extra employee. It takes over the retyping; your team keeps the profession.
A specific question about your obligations?
AMB HOLDING SARL uses this data to contact you back and measure the effectiveness of our ads. Data & rights
Or reach us directly