Skip to main content
This page contains a set of frequently asked questions regarding the safety and security of Legacy Pagos AI. As a Beta product, we actively monitor Pagos AI’s performance and welcome your questions and feedback. Please contact us if:
  • Information appears incorrect compared to your analyses or our documentation
  • You have questions about how Pagos AI generated a specific result
  • You’d like to disable access to specific data sets

​
Data and Timing FAQs

​
What data is Pagos AI trained on?

Pagos AI is not separately trained. It uses a standard ChatGPT-based assistant (provided by OpenAI) configured with:
  • Pagos Product Documentation (provided as a searchable resource)
  • Access to query your anonymized payments data through our API
  • System prompts that define its intended behavior
Pagos doesn’t perform any additional model training, and our contract with OpenAI prevents them from using sessions for training their models.

​
Does Pagos AI have access to my organization’s data?

Pagos AI can only access aggregated payments data based on what you have already provided to the Pagos platform through your data connections. By default, your organization is not directly identified in queries, except:
  • If you include identifying information in your prompts
  • If you query MIDs associated with your business (merchant names may contain your organization’s name)
You can request that we disable access to your MID data at any time.

​
Accuracy & Reliability FAQs

​
How accurate is Pagos AI?

Pagos AI is constrained to answer only payments-related questions using:
  • Your payments data on the Pagos platform
  • The Pagos Product Documentation
  • General payments industry information
Through extensive testing, we’ve rarely found discrepancies. AI can occasionally generate imperfect information, however, so we recommend verifying important technical insights against your own analysis on the Pagos platform or with offline tools.

​
How does Pagos AI handle incorrect information or “hallucinations?”

Our system is designed to keep Pagos AI tightly focused on payments questions related to your data and our documentation. The assistant is configured to state when it can’t reliably answer a question. As a Beta product, we encourage caution when interpreting responses and welcome feedback to help us improve.

​
Can I verify how Pagos AI generated its results?

Yes. You can ask Pagos AI to explain:
  • Its query logic (which data fields, filters, or groupings it used)
  • What methods or resources it used to determine an answer
  • Citations or code snippets supporting its response
For some analyses, you can also request a Python script to run externally on exported data for independent verification.

​
Safeguard FAQs

​
What safeguards does Pagos AI have in place?

Pagos AI operates within strict safety boundaries. Pagos maintains complete control over the source data used for interactive sessions and data from your sessions are not used to train AI models. Pagos AI:
  • Only provides information based on your specific requests
  • Is constrained to answer payments-related questions only
  • Has read-only access to your data within a sandbox environment
  • Doesn’t share any transaction-level data with ChatGPT
  • Doesn’t share information identifying customers or your organization with ChatGPT
  • Can’t modify any application settings
  • Can be instructed to alter its analysis approach within its safety limitations