Proceed with Caution: AI Transcription Tools May Increase the SAR Workload
There is no doubt that the global AI transcription market is on a stellar path, with many reports claiming it is set to grow 14.5% by 2027, with the overall speech and voice recognition market reaching $31.8 billion by 2025.
But with that growth comes extra risks, particularly when data subject access requests (SARs) are added to the mix because all transcript documents and data records created using AI assistant tools – such as those found in Teams or Zoom – must be included in the SAR response.
A potential minefield
Anyone hosting conference calls must get consent from all participants before using AI assistant tools to record. Most meeting hosts do request consent, but many AI transcription tools on the market can also transcribe meetings that have been covertly recorded—using a phone, for example—allowing individuals to document ‘off-the-record’ calls or meetings.
This could cause significant repercussions for companies involved further down the line when a SAR request is actioned and all this unexpected and unpoliced material is suddenly included in the completed DSAR.
Also, those companies that fail to ensure the right internal policies are in place when it comes to the creation of transcripts and summaries, or fail to implement strict rules about information/document sharing and indeed file deletion internally, could be opening themselves up to a world of pain if a wider regulatory investigation is triggered as-a-result of a poorly-handled SAR.
What can go wrong?
Some AI tools will give the option of creating meeting summaries or action points but could fail to capture the meeting content fully. Context could also be lost/missed, and the subtle nuances of a conversation could be misinterpreted because the AI tools are just not as advanced as a human transcriber.
In addition, the accuracy of some AI transcription tools due to language differences, accents, multiple speakers, or even background noise could come into question. Mistakes will be made, and the wrong people could be quoted saying something they didn’t actually say. Or random words could be added to a conversation because the technology failed to capture to actual word used.
Some of the words misused or completely made up could be trigger words for a SAR, and it could end up costing a company not only financially (because someone will have to go through all the material collected and find out where the error initially occurred), but also reputationally.
Also, if a meeting is being recorded, it is possible that the AI transcription technology could capture ‘off-the-cuff’ conversations—for instance, during a hiring process between HR and the hiring manager—which were never meant to be documented or shared. If these informal remarks include personal opinions, sensitive information, or references to a candidate by name, they could inadvertently create additional disclosable material in the context of a Subject Access Request. This not only increases the volume of data that needs to be reviewed but also introduces potential risks around confidentiality, fairness, and compliance.
How Smartbox.ai can help you safely combine AI transcription and DSARs
By enlisting the help of experts such as Smartbox.ai to take the SAR headache away and ensure correct data management practices are upheld, management teams can rest easier at night.
Smartbox.ai technology will do the hard work, collating all relevant information and enabling the review and redaction of single assets, a group of assets, or a complete data set before it falls into the wrong hands.
Smartbox.ai technology can also identify duplication in documents and email threads, and use risk indicators set by the client to highlight any trigger words in a dataset and navigate easily to problem areas. Find out more about Smartbox.ai and request a demo: https://smartbox.lislex.xyz/dsar-management