Social media site LinkedIn has been accused of collecting data from unwitting Australian users to train its artificial intelligence “models” without telling them.
Creative media expert Dr James Birt has criticised the tech giant for forcing users to opt out of a policy they didn’t even know existed, after automatically setting their accounts to opt in to having their profile data plundered.
“This shows the dark side of Big Tech,” said Bond University associate professor He told news.com.au.
‘While users can opt out, the setting is enabled by default, raising concerns about informed consent.
‘The lack of proactive uptake is a typical example of how big tech companies exploit user apathy or lack of awareness to promote their AI initiatives.’
The setting, known as “Data for Generative AI Improvement,” has been automatically turned on for users outside of the EU, EEA, UK or Switzerland, giving LinkedIn and anonymous “affiliates” permission to “use your personal data and the content you create… to train content creation AI models.”
Leaving it on allows the app to share users’ activities with its anonymous “affiliates” in order to “train content creation AI models.”
LinkedIn has been criticised for automatically enabling a preference in the app that allows users’ data to be mined for use in its artificial intelligence systems.
Anyone using the app who does not want LinkedIn to record their activity must manually disable ‘Data for Generative AI Improvement’.
This includes anything someone posts or even the content of a user’s profile.
Turning off this setting will prevent LinkedIn from collecting data in the future, but it will not delete data that LinkedIn has already collected while the setting was active.
LinkedIn is not the only popular app that “mines” user data for its own benefit.
Meta, the company behind Facebook and Instagram, confirmed earlier this month that it had also stored data on photos and posts from Australian adult users dating back to 2007.
Meta made the admission when her Privacy policy director Melinda Claybaugh appeared before an inquiry and said the company collected data under pressure from senators.
Meta, which owns Facebook and Instagram, confirmed in early September that it had stored data on photos and posts from Australian adult users dating back to 2007. Meta made the admission when its privacy policy director, Melinda Claybaugh, appeared before a commission of inquiry.
Regulations surrounding this type of behavior have been tightened in Europe, where companies need permission to store user data, but in other areas such as Australia this courtesy is not required.
Dr. Birt said the decision to automatically include users in this type of practice exemplifies the ethical concerns around storage of personal data.
LinkedIn claims it uses Generative AI for “a variety of purposes,” including its writing assistant that helps users compose messages.
Microsoft owns the platform and these AI models are on LinkedIn, but not on Microsoft’s Azure OpenAI, which created ChatGPT.
LinkedIn Spokesperson Greg Snapper clarified that the app was “not sending data to OpenAI to train their models.”
Leaving the setting on allows the app to share users’ activities with its anonymous “affiliates” in order to “train AI models for content creation.”
Daily Mail Australia has contacted LinkedIn for comment.
To disable the feature in the app, users must tap on their profile, go to settings, then data privacy, and finally access “data for generative AI improvements.”
From there you can disable the setting.
When users click on ‘‘Learn more’ In the final stage, the app explains its use of AI.
‘This setting applies to the training and tuning of generative AI models used to generate content (e.g., suggested posts or messages) and does not apply to the development of AI models used by LinkedIn or its affiliates for other purposes, such as models used to personalize your LinkedIn experience or models used for safety, trust, or anti-abuse purposes,’ it states.
On LinkedIn Generative AI FAQ: The app states that it will “attempt to minimize personal data in the datasets used to train the models.”