ALBAWABA - Meta has updated a section in Facebook's Help Center on its website this week to include a feature titled "Rights for Data Used in AI Training." This feature allows users to "submit requests regarding your third-party data used in training generative AI models."
The company is adding an opt-out tool with the rollout of generative AI technology, where companies create more advanced chatbots and transform simple text into sophisticated responses and images. Meta gives individuals the option to access, modify, or delete any personal data included in various third-party data sources that the company uses to train its extensive language and AI models.
In the feature, Meta refers to third-party information as data "publicly available on the internet or licensed sources." The company notes that this type of data can represent some of the "billions of pieces of data" used to train generative AI models that "use predictions and patterns to create new content."
In a related blog post on how data is used for generative AI, Meta states that it collects publicly available information on the web in addition to data licensed from other service providers. Meta said that blog posts, for instance, might contain personal information like someone's name and contact information.
The model does not take into account a user's activity on Meta-owned platforms, whether it's Facebook comments or Instagram photos, so the company could use this first-party data to train its own generative AI models.
A spokesperson for Meta said that the company's latest open-source language model, Llama 2, "has not been trained on Meta user data, and we have not launched any consumer-facing features that leverage generative AI on our systems yet."
The spokesperson added: "Depending on where people live, they may have the ability to exercise their data rights and object to the use of specific data for training our generative AI models," referring to various data privacy regulations outside of the United States that give consumers more control over their personal data used by technology companies.
Like many of its tech peers, including Microsoft, OpenAI, and Alphabet, the parent company of Google, Meta gathers vast amounts of third-party data to train its models and related AI programs.
Meta stated in the blog post, "To train effective models that unlock these advances, a significant amount of information is needed from publicly available and licensed sources." The company also added that "using public information and licensed data is in our interest, and we're committed to being transparent about the legal bases we use to process this information."
However, in recent times, some data privacy advocates have raised concerns about the massive collection of publicly available information for training generative AI models.
Last week, a group of data protection agencies from the UK, Canada, Switzerland, and other countries issued a joint statement to Meta, Alphabet, TikTok's parent company ByteDance, X (formerly known as Twitter), Microsoft, and others regarding data collection and user privacy.
The goal of the message was to remind social media and technology companies that they remain subject to various data protection and privacy laws worldwide and that "they protect the personal information that can be accessed on their websites from exploitation."
The group said in the statement, "Individuals can also take steps to protect their personal information from theft, and social media companies have a role to play in enabling users to interact with their services in a privacy-protective manner."
Here's how to delete some of your Facebook data used for training generative AI models:
Go to the "Rights for Data Used in AI Training" feature on Meta's Privacy Policy page about generative AI.
Click on the link that says "Learn more and submit requests here."
Choose from the three options where Meta says it "best describes your problem or objection."
The first option allows individuals to access or download any of their personal information that was collected from external sources used for training generative AI models. By selecting the second option, they can delete any of their personal information from third-party data sources used for training. The third option is for individuals with a "different problem."
After selecting one of the three options, users will need to pass a security check. Some users have reported being unable to complete the form due to what appears to be a programming error.
