TLDR: As generative AI models increasingly rely on extensive datasets for training, tech companies are revising their privacy policies, sparking significant user concerns regarding data privacy. While some platforms offer options to opt-out, the effectiveness of these measures can vary, and data already used for training may not be removed. The article highlights methods for individuals to attempt to prevent their online data from being utilized for AI training.
The rapid expansion of generative AI models, which are fueled by immense quantities of data, has prompted numerous technology companies to update their privacy policies and terms of service. These revisions often permit the use of user data for training AI models or licensing it to other entities for similar purposes, leading to growing apprehension among users about their data privacy.
A recent incident involving the file-sharing service WeTransfer exemplifies this concern. The company faced immediate user backlash after it amended its terms of service to suggest that uploaded files could be used to ‘improve machine learning models.’ Although WeTransfer subsequently removed any mention of AI and machine learning from the document, this event underscored the intensifying user concerns over privacy and data ownership in the AI era.
Tech companies are actively scraping publicly available, copyright-protected data from across the internet to train their AI models. This includes a wide array of online content, from social media posts and blog entries to restaurant reviews and personal photos. While this indiscriminate data scraping has led to legal challenges from artists, content creators, and other rights holders, individual users also have avenues to explore to prevent their online contributions from being used for AI training.
In response to mounting user pressure, many companies now provide individuals and business customers with options to opt-out of having their content used in AI training or sold for such purposes. For instance, users concerned about their work being scraped for AI training can visit ‘Have I Been Trained?’, a service operated by the tech startup Spawning. Specific platforms also offer opt-out mechanisms:
LinkedIn: Users can toggle off the option labeled ‘Use my data for training content creation AI models.’
ChatGPT and DALL-E (OpenAI): For its image generator DALL-E, OpenAI has stated that users wishing to remove their images from future training datasets must submit a form with their details, including name, email, and confirmation of content ownership.
Also Read:
- WeTransfer’s Terms of Service Backtrack Highlights AI, Data, and Digital Trust Challenges
- SoundCloud Revises AI Policy Amidst Artist Concerns, Emphasizing Support Over Replacement
It is important to note that while these opt-out options are becoming more prevalent, choosing to opt-out does not necessarily guarantee that data already incorporated into AI training or existing datasets will be erased.
