FortiAI data privacy

FortiManager and FortiAI protects your data by masking private information such as IP addresses before it is sent to the FortiAI large language model (LLM) for processing. In this topic you can find a list of protected data as well as the process FortiManager follows to protect your data.

Protected data

The following list of data is considered private and will be masked on FortiManager before it is sent to the FortiAI LLM. See How private data is protected.

• IoT devices' MAC addresses, vendors, and hostnames

• The FortiGate device name on FortiManager

• The "root" keyword

• VDOM names

• IPv4 and IPv6 addresses

• MAC addresses

Private data included in images such as topologies that are uploaded to FortiAI will not be masked when the image is sent to the LLM for processing. When uploading an image to FortiAI, FortiManager will present a warning message that the administrator can use to confirm or cancel the upload before it is sent to the LLM for processing.

How private data is protected

1. The FortiAI assistant identifies information in a query that matches the list of protected data.

2. FortiManager masks the private data, and the masked data is returned to the FortiAI assistant.

3. The FortiAI assistant creates a one-to-one mapping between the masked and unmasked data.

4. The FortiAI assistant sends the masked data to the LLM where the request is processed.

5. When the result is returned, FortiAI receives the masked data from the LLM, and a reverse mapping is performed.

6. The private data is returned to the user unmasked in the assistant's response.