LinkedIn trains generative AI models on user data, offers opt-out option to address privacy concerns

In what could be described as an embarrassment to LinkedIn, the social networking site for professionals was reportedly scraping user data on the platform without informing the users first.

The business and employment-focused social media platform that works through websites and mobile apps has now updated its terms of service to reflect this, however, it continues to automatically opt-in users unless they manually find and toggle off the setting to opt-out.

Several netizens took to social media platforms to criticize the move by the company. Pertinently, the updated policy highlights that the Artificial Intelligence (AI) is being trained for features such as writing suggestions and posting recommendations.

How LinkedIn got exposed

A 404 Media report highlighted LinkedIn's data scraping before the company updated the policy informing users about its action. Many users also posted on the platform on Sep 18 claiming they found a new option in settings about data training for AI.

Training AI models on user data is not a new practice. Earlier, Meta admitted to training its in-house Llama models on publicly available user posts, and Google updated its policy last year to train Gemini and other AI models on publicly available web data.
However, the unlike other social networking sites, LinkedIn started collecting user data before it publicly informed users.'

The social media sites are expected to share a prior intimation to users about such decisions is to enable them to proactively opt-out and protect their privacy. LinkedIn's move to not follow this protocol has led netizens to take to social media platforms to criticize the move.

However, LinkedIn has now updated the policy to announce the change. It now explains that “LinkedIn or its affiliates train or fine-tune generative AI models used to create content, including content that may be distributed or made available on LinkedIn's platform”. The Microsoft-owned company also revealed that the training was done for AI-powered writing assistants as well as post-recommendation features.

LinkedIn was earlier quoted by TechCrunch - a news website dedicated to the tech scene- as saying, “The company used privacy-enhancing techniques to limit the collection of personal information, which was used to train generative AI models. Some of the measures included redacting and removing personal and identifying information,”

To prevent LinkedIn from scraping data on the platform, users can go to the settings option and choose off option for “Use my data for training content creation AI models”. However, it is still not clear as of now whether the social media platform will also remove all the previously collected data from the AI models' datasets.