ASR transcription with data privacy
Feb 17, 2022

How ASR Delivers Data Privacy for Transcription Services

Data privacy is an enormous concern for consumers wondering who may be viewing their data and the possibility that it can be leaked. When data is stored in a shared environment or human workers are viewing it, the risks are even greater. For traditional transcription services, these risks are part of standard practices. Recently, issues relating to privacy have surfaced with Facebook’s transcription audio scandalGoogle’s data leak, and leaked voicemails with confidential information.

While banking, healthcare, legal, and contact centers are obvious areas where protecting personally identifiable information is an imperative, any company that receives sensitive user information should be aware of its vulnerability to data leaks and exposure through traditional transcription services. Now more than ever, consumers are demanding transparency from companies regarding how their information is being collected and protected, and companies need to address these issues if they hope to avoid the negative impacts of data leaks. 

As a safer alternative to traditional transcription methods, automatic speech recognition (ASR) technology solutions remove the human element and shared data storage from the transcription equation—keeping personally identifiable information secure while providing fast and efficient transcription services. By using ASR technology to transcribe everything from legal documents and patient conversations to banking customer service calls, consumers don’t have to worry about who is seeing their personally identifiable information, if it’s getting leaked, or if their voiceprint is vulnerable to hacking.

Let’s examine these 3 questions about ASR and transcription services:

  1. Why is data privacy essential for transcription services? 
  2. How do ASR solutions protect data privacy for transcriptions?
  3. What industries are poised to use ASR for transcriptions? 

Why data privacy is essential for transcription services

Transcription services are essential for companies that need to keep written documents of conversations. Transcriptions help with record-keeping, training, accessibility, comprehension, and preserving important documents, but they also come with their own set of concerns about data privacy, especially for personally identifiable information.

Personally identifiable information can consist of direct information, such as social security numbers, driver’s license numbers, financial or medical records, addresses, phone numbers, and email addresses, or indirect information like a combination of gender, race, birth date, geographic location, and other personal descriptors.

Traditional transcription services are vulnerable to privacy concerns due to a few key factors, including: 

  • Human workers viewing personally identifiable information
  • Shared environments and workflows that are not properly encrypted
  • Identifying identities based on voiceprints

When people are handling sensitive information, companies have to consider the possibility that data can be misused and leaked. Many transcription services are located offshore, which have different, often less strict, data privacy laws than in the US or EU. The companies using these services aren’t managing the transcribers directly, have no way of knowing exactly how personally identifiable information is being handled, and aren’t aware of the vetting process for employees.

Many transcription services are located offshore, which have different, often less strict, data privacy laws than in the US or EU.

Some companies will attempt to identify and redact the personally identifiable information before it reaches human workers, but it’s not a perfect system since the information can be very difficult to spot beforehand. In some cases, companies will attempt to simply beep out the personally identifiable information in the audio so that it won’t appear in the transcription at all. 

Companies have also tried breaking down the audio into small 20 second segments for each person to increase data privacy. However, this often creates issues with inconsistency in the transcription. The practice is labor-intensive and only somewhat effective as there will still be a human at the end of the process who proofreads the entire clip. When there are people involved, true data privacy can’t be completely achieved. 

When there are people involved, true data privacy can’t be completely achieved. 

Another major concern for data privacy in transcriptions is the environment in which the data is being stored and transmitted. If the data is being stored in a shared environment, there is a much greater chance that the data could be hacked and leaked. For the best data privacy, storage needs to be in a dedicated, isolated environment. Also, companies need to consider if the data is being properly encrypted at every level; otherwise, it could be intercepted. 

Some transcription services may also store the audio files for long after the transcription has been completed, sparking concerns over whether or not speakers could be identified solely on their voiceprint. While standard practice is to destroy the files within 3 years, it wouldn’t take that long for the data to be leaked or intercepted if it’s not being properly stored or encrypted.

How ASR solutions protect data privacy for transcriptions

When considering transcription services, companies should consider these essential questions around data privacy:

  • Are the files properly encrypted?
  • Can any voices be identified? 
  • Will the transcription vendor keep the voice data? 
  • What’s being done to change the identity of the voice? 
  • Who is seeing the data?
  • If a person is in the loop, where are they based, and what are the data privacy laws? 
  • Is personally identifiable information being redacted? 

When answered from the perspective of traditional transcription services, the responses would reveal serious concerns about data privacy. In contrast, advanced ASR technology addresses each of these areas with solutions that deliver greater safety of personally identifiable information. 

Voice-enabled transcription services eliminate the privacy concerns around traditional transcription methods by properly encrypting the files in dedicated environments, ensuring no humans see the data, and using voice morphing to mask the identities of the speakers. 

Voice-enabled transcription services encrypt files in dedicated environments, ensure no humans see the data, and use voice morphing to mask the identities of the speakers. 

With ASR technology, consumers don’t have to worry about who is seeing their data or unintended uses of their personal identity. Everything is handled by an AI, ensuring that no human will have access to their personally identifiable information. Companies don’t need to look into where the workers are based, what the privacy laws are, or whether or not the information is redacted. Everything can be transcribed safely, quickly, and efficiently.  

Voice AI technology also is developing voice morphing to change the customer’s voice to the point where the voice can’t be identified or stolen, removing any ability to associate a voice with a person. With the added anonymity, companies can then use the data to train the system without any privacy concerns. Data is essential to train and continuously improve voice models. If the identity is masked with voice morphing, then the data can still be used after the transcription to further improve the voice assistant while keeping the identity safe and secure.  

What industries are poised to use ASR for transcription

While any industry or company with personally identifiable information can benefit from voice AI transcription services, the following industries are especially poised to use voice AI due to the amount of sensitive information they handle: 

  • Healthcare
  • Legal
  • Banking
  • Contact centers

Transcriptions of patient interactions can be hugely helpful for record-keeping and future review, but the information is highly confidential. Already, the medical company, MEDantex, faced a data leak of personally identifiable information from the transcription service they used. ASR technology adds a level of security and safety by transcribing the audio and eliminating concerns over data privacy.

The medical company, MEDantex, faced a data leak of personally identifiable information from the transcription service they used. 

Contact centers, from retail to banking, use transcriptions for a variety of purposes. They help preserve calls for future training, help superiors advise a representative on how to better handle a call, and keep records of what was discussed. However, a lot of personally identifiable information is also transmitted on those calls from credit cards to social security numbers. 

Another reason voice AI transcriptions are optimal for contact centers is that the calls happen in real-time, so the process has to be automated. Voice assistants can transcribe as the call is happening while keeping sensitive information protected. 

Data privacy is more important than ever, and companies looking to preserve their relationships with their users need to ensure customer data is being protected. Traditional transcription services are vulnerable to a variety of security risks, including human workers, shared environments, and identifiable voiceprints. Advanced ASR solutions, like those offered by SoundHound, eliminate those concerns by providing safe, fast, and efficient transcription. 

At SoundHound Inc., we have all the tools and expertise needed to create custom voice assistants and a consistent brand voice. Explore SoundHound’s independent voice AI platform at and register for a free account here. Want to learn more? Talk to us about how we can help bring your voice strategy to life.

Kristen is a content writer with a passion for storytelling and marketing. When she’s not writing, she’s hiking, reading, and spending time with her nieces and nephew.

Interested in Learning More?

Subscribe today to stay informed and get regular updates from SoundHound Inc.

Subscription Form Horizontal