How the Assistant uses AI
Probe Software Assistant is an AI system, not a human analyst. This page explains how it builds answers and where information is processed.
How answers are generated
The Assistant combines automated search and retrieval from Probe Software's curated knowledge sources with language models supplied by an external AI service provider. It may also use saved conversation context and an instrument profile when those details are relevant.
Probe Software may change the underlying model, routing, retrieval method, or provider as the Service evolves. We do not publish a model identifier in the normal user interface because the product is the complete support workflow, not a promise to use one fixed model.
What may be sent through the AI API
When you ask a question, the Service may transmit the following information to the AI provider through an application programming interface, or API:
- your current question;
- relevant portions of the current conversation;
- selected excerpts from the knowledge base and internal response instructions;
- relevant instrument or material context; and
- an image you intentionally attach when image analysis is available.
You are not sending information directly from your browser to OpenAI. Probe Software's server makes the API request, but the included content is still transmitted to and processed by OpenAI to produce the response.
Do not submit passwords, private keys, patient data, controlled technical data, confidential client information, unpublished trade secrets, or other sensitive information unless you are authorized to disclose it and your organization permits the use of this Service for that information.
OpenAI data handling
The Service currently uses OpenAI language models through the OpenAI API. Under OpenAI's current API data policy, API data is not used to train OpenAI models unless the API customer affirmatively opts in. OpenAI may retain prompts and responses in abuse-monitoring logs for up to 30 days by default, and may retain information longer when required for safety or legal reasons.
Provider practices can change. See OpenAI's current API data controls and retention documentation, and read our Privacy Notice for the broader handling of account and Service information.
What you must verify
AI output can be inaccurate, incomplete, outdated, internally inconsistent, or mismatched to your instrument, material, phase, analytical goal, or software version. A confident tone, source link, calculation, or detailed setup does not prove an answer is correct.
Before acting, verify the original source and empirically confirm analytical settings, element and line choices, crystal and spectrometer compatibility, standards, peaks, backgrounds, interferences, beam conditions, calculations, uncertainty, and safety requirements. The Assistant is not an autonomous instrument operator, an accredited procedure, or a substitute for qualified professional judgment.