‍

A fascinating report published by the PMA and University of Edinburgh, ‘PSM and AI’, has revealed that AI transcription has public service media professionals nervous around data privacy risks. The report suggests that journalists are often using cheap or even free AI transcription tools, which raises serious concerns about where confidential information might end up.

‍

The good news is that not all AI transcription solutions are dangerous for data privacy. Choosing a provider that has taken the necessary security measures can ensure your organization benefits from AI transcription without the risk.

‍

‍

In a nutshell, many popular AI transcription tools are not built with security and data privacy in mind. In other words, they’re not secure by design, and this has led to a couple of major concerns:

‍

Their AI is training on your data

While for the purpose of improving the accuracy of their AI models, this presents huge implications for the integrity of your data - indeed compromising the confidentiality of your stories and sources. For newsrooms, this could undermine your editorial independence and, critically, public trust.

‍

The report cites a prime example of where Australian Broadcasting Corporation (ABC) had banned a high profile AI transcription vendor (see p.12 of the report) after failing its data privacy and security assessments. The vendor in question is known for using customer data to train its AI, while their software also automatically joins, records and shares meeting transcripts without the user’s knowledge. 

‍

They store data only in one geographical region

Many AI transcription solutions store their data only in the US, which potentially exposes media organizations to GDPR compliance risks and potential legal challenges, particularly if they’re handling sensitive or confidential information related to EU citizens.

‍

‍

To avoid these risks, journalists and news organizations should look for AI transcription providers that are secure by design, prioritizing the privacy of customer data.

‍

Here's a checklist of features to look for in a safe AI transcription solution:

‍

1. No training on your data. The solution should explicitly state in their privacy (or equivalent) policy that it does not use your transcripts to train its algorithms. At Trint, we use external data sets to train our algorithms, meaning no one sees your data but you. In sum, your sensitive information remains private.

‍

2. GDPR compliance. Even if your organization is not based in Europe, choosing a GDPR-compliant provider is your most effective path to ensuring best practice data protection. With Trint, customers have a choice of storing data in the US and EU, while you have clear data retention and secure deletion capabilities at your disposal. That way, as the report highlights, you can rest assured that Trint is fully GDPR compliant. Your data remains within the EU for a start, while you have greater control and peace of mind regarding the sovereignty of your data.

‍

3. Robust encryption. All data, both in transit and at rest, should be encrypted using industry-standard protocols. Trint encrypts all data in transit using TLS 1.2 or later and at rest with AES 256-bit encryption. That way, you minimize the risk of unauthorized access while maintaining data integrity. 

‍

4. Secure Infrastructure. Your solution must use secure, certified data centers with strong physical and digital security measures, as well as best-in-class resilience should the worst ever happen. Trint runs on AWS S3 so, when users work in Trint, their content is immediately backed up into three availability zones with a durability of 99.99%, and availability of 99.99% over a given year. In short, you’re in good hands.

‍

‍

AI transcription offers great advantages of speed and efficiency, but it feels like too many vendors are compromising data privacy - and newsrooms are increasingly becoming wise to the risks.

‍

Choosing the cheap (or even free!) option might look great on the surface, saving you and the organization money - but the risks to data and journalist sources are grave. However, by choosing a solution like Trint that has been built from the ground up with security and data privacy firmly in mind, newsrooms can confidently enjoy the benefits of  AI transcription without compromising their core values of trust and accountability.

‍

‍Read the full ‘PSM and AI’ report for more about how public media organizations are approaching AI. Or if you’re interested in exploring Trint’s AI transcription for your newsroom, with security and GDPR compliance in mind, speak to our team for a live demo.