Paytia lets AI agents accept cards without exposing data

Paytia launched Conversational & AI Payments to let phone and chat AI agents take card and bank payments while keeping card numbers out of AI networks, transcripts and logs.

Paytia announced Conversational & AI Payments this week, a service that lets AI phone and chat agents accept card and bank payments while keeping raw card numbers and other sensitive fields outside the AI systems, transcripts and logs.

When an AI agent reaches payment, it calls Paytia’s Capture Assist API. Customers enter card details on a Paytia-hosted web form or by keypad during a call. Paytia processes the data on its SecureFlow PCI DSS Level 1 platform, returns a token and a result, and the bot resumes the conversation without ever receiving the raw card number.

The same isolation can capture bank account numbers, passport numbers and national insurance numbers when required. For voice agents Paytia connects over SIP in two modes: it can sit in front of the bot to prevent sensitive data entering the bot’s environment, or it can be conferenced in on demand using a unique call ID in the SIP header to trigger capture. Chat agents use Advanced Payment Links to collect payment details.

Paytia says the service works with existing platforms and that current Paytia customers can enable Conversational & AI Payments through a configuration change rather than a new contract. The company has published a Data Protection Agreement that defines how captured data is handled, retained and deleted and acts as the appointed data processor for customers.

Curtis Nash, Paytia’s chief executive, commented: “A card number should never sit in a chat log or an LLM’s context. The bot runs the conversation, we hold the sensitive data, and the two never mix.”

Paytia holds PCI DSS Level 1 and Cyber Essentials Plus certification, reports it has processed more than £400 million since 2020, and says Conversational & AI Payments are available now. The company also states that card data is not logged or used to train models.

Payments industry rules require strict controls over cardholder data. Paytia’s product uses tokenization so agents can reference payment results without storing card details, aiming to keep payment processing and conversational AI systems separate.

Articles by this author