Excessive free comments on a radio station's listeners
In 2016, the CNIL looked into in the comment databases relating to the listeners of a radio station These include:
- excessive comments and insults
- data relating to assumed racial origin ("Tunisian Jewish accent", "North African accent")
- health data
- data related to sexual orientation
In addition to their problematic nature, some of the open-ended comments date back to 2002. The company actively and voluntarily cooperated with the regulator, putting an end to these practices in 2017. Therefore, there was no public sanction. On the other hand, again, there was a significant media fallout from this case.
HRD - Keep an eye on your employee tracking files and your free comment areas
The management of personnel, temporary employees or trainees and recruitment campaigns can result in the recording of tracking files. These files contain very useful information but it is also an opportunity for their authors to "let loose."
These various companies stopped the practices involved and were forced to publicly apologize.
Customer relationship management can get out of hand because of free comments
Some e-commerce sites have fallen victim to bad practices. Customer relationship managers can be tempted to enter derogatory or even illegal comments in the CRM for all sorts of reasons.
- In 2015, an e-commerce site received a public notice from the CNIL. The presence of illicit comments was highly publicized. This case was all the more publicized because the CNIL reproached the company for numerous failures regarding the GDPR compliance of its site, overall.
- The same year, a dating site was also the subject of a public notice, notably because of comments qualifying customers ("ball", "whore")
- In 2016, another e-commerce site was subject to a public warning by the CNIL
What about you? Have you brought your free comment areas into GDB compliance?
Monitor your free comment areas with Deep Learning
How can you control the use of free fields in your CRM, HRIS or ERP and moderate comments? You can, for example, create a dictionary of keywords that an automated tool will be able to prohibit from being entered. However, this dictionary must be continually enriched, as this method is often incomplete when faced with ambiguous and graphic expressions. Lettria has chosen a hybrid methodology that combines dictionaries of prohibited words, "pattern" detection of problematic sentences, and a classification model trained on our proprietary data. Only these combined techniques can apprehend the richness of the spoken language, and to correctly detect offensive comments, as well as sensitive personal data.