AI Call Scoring & AI Call Summary: How Do My Call Transcripts Interact With The AI & Is It Safe?

If you have ever used OpenAI's ChatGPT, Microsoft Copilot, Google Bard, or any other Artificial Intelligence (AI), you might have asked yourself: are my chats used to train the model? Are they saved somewhere? Will the information I provide be accessible to someone else? 

Here at MiaRec, we have not only developed our own proprietary (and much more affordable but just as accurate) speech recognition engine using AI almost ten years ago, but we also have helped hundreds of organizations around the globe increase compliance and keep their data safe. So, we know data security and privacy are of utmost importance. 

In this article, we will explain how MiaRec's Auto Score Card and Auto Call Summary interact with the AI, which data is send the the Large Language Models, and we will address any compliance concerns. By the end of the article, you will understand why this isn't any different from saving files on DropBox or interacting with any cloud service as the same industry security standards apply.

Side Note: For many, this stems from the initial terms and conditions of OpenAI's ChatGPT, stating that they will use your AI interactions to train their model. However, the company has since removed the clause from its paid services. It is essential to understand that if you use commercial services from OpenAI, they don't use your data for training. In other words, nobody reads your conversations or uses your data in training. 

How Do MiaRec's Auto Call Summary & Auto Score Card Interact With AI?

Before diving into how secure everything is, we need to understand how exactly products interact with AI. MiaRec's Automated QM and AI Assist products heavily rely on AI.

MiaRec Auto Call Scoring uses Generative AI to automatically score all of your calls by "reading" the whole transcript (which is also transcribed using AI) and then evaluating the agents' performance based on criteria and questions you defined. Auto Call Summary will also read the transcript of your call recordings and summarize the conversation based on the parameters you provided. For example, you can give instructions to make it very short and always extract specific key facts or make it more comprehensive. 

Without getting technical, you must know that MiaRec submits the transcript, customer instructions, questions, etc., to the Large Language Model (LLM). The Generative AI will follow the instructions, process your transcripts according to them, and return the results. 

In the industry, technology vendors have to encrypt data in two ways: in transit and at rest. This means your data is sent to the AI encrypted, where it is temporarily stored (also encrypted), and then, after it serves its purpose, it is deleted. Depending on the specific provider's terms, this can be a few days up to 30 days at your request. Remember: holding this amount of data is very expensive for any technology provider, so shorter storage times are better for them.

I Need To Be PCI-DSS or HIPAA Compliant. Can I Redact Sensitive Data Before Sending Transcripts?

The short answer is: Yes, you can redact your MiaRec transcripts before sending them to the AI for processing. However, it is essential to understand that each regulation's data security and privacy requirements differ.

If you are worried about PCI-DSS compliance, use MiaRec's Auto Data Redaction feature, which automatically redacts sensitive data such as credit card numbers, security codes, and other payment card information in the audio file and the transcript. This data is not recoverable. 

However, for HIPAA, the situation is entirely different. You may store sensitive data such as the patient's name and their diagnostic data, but it must be encrypted, which makes sense because if you redact it, the transcripts would be useless nonsense. To accomplish this, every technology vendor has to sign an agreement to be HIPAA compliant by providing some guarantees that they don't use the data.

Essentially, both sides promise not to look at the data. If you are a healthcare organization using Amazon Web Services, you have an agreement with them. The same is true for MiaRec. We send the data encrypted to the LLM and contract with the AI provider, just as we deal with Amazon Web Services or other cloud services.

Conclusion

In summary, MiaRec's Auto Call Summary and Auto Score Card utilize Generative AI and Large Language Models. The to-be-scored or -summarized transcripts are send to the AI together with any custom instructions and scorecard questions or evaluation criteria. To ensure that your information is handled with the highest level of protection, MiaRec follows all industry data security and privacy standards by encrypting data in transit and at rest. Whether you need to comply with PCI-DSS or HIPAA regulations, MiaRec offers features like Auto Data Redaction to safeguard sensitive information.

To experience the seamless integration of AI firsthand and explore how MiaRec can benefit your organization, we encourage you to book a demo and witness the efficiency and reliability of our solutions in action.