When the Terms of Service Change to Make Way for A.I. Training

🌈 Abstract

The article discusses how tech companies are modifying their privacy policies and terms of service to allow the use of public and private data to train their AI models, which is raising concerns among some users, particularly creators, about their content being used without their consent.

🙋 Q&A

[01] Changing Privacy Policies and Terms of Service

1. What are some examples of how companies are modifying their privacy policies and terms of service to enable AI training?

• Google updated its privacy policy to state that it uses publicly available information to train its language AI models and build products like Google Translate, Bard, and Cloud AI capabilities.
• Snap updated its privacy policy to explain how its AI chatbot "My AI" uses the information collected from user interactions to improve Snap's products and personalize the user experience.
• X (the article doesn't specify the company) added a single sentence to its privacy policy about using information to train its machine learning or AI models.
• Meta alerted its Facebook and Instagram users in Europe that it would use publicly available posts to train its AI, though it later paused these plans after receiving complaints.

2. How are some companies trying to protect their content from being used to train competing AI systems?

• Adobe added language to its terms of service prohibiting users from using its services or content to directly or indirectly create, train, test, or improve any machine learning algorithms or AI systems.

3. What are the concerns expressed by some users, particularly creators, about their content being used to train AI without their consent?

• Some users, like Sasha Yanshin, a YouTube personality and co-founder of a travel recommendation site, are worried that their work is being used to train AI products that could replace them.
• There are concerns that entire segments of the creative industry could be "decimated" as AI-generated content competes with human-created content.

[02] Accessing Private Data for AI Training

1. What are the two main sources of data that tech companies can use to train their AI models?

• Public data, which is available on the web for anyone to see
• Private data, such as text messages, emails, and social media posts from private accounts

2. What are the challenges in accessing private data for AI training?

• Private data is mostly protected by a patchwork of federal and state privacy laws that give users some licensing over the content they create online.
• Companies cannot use private data for their own products without user consent.
• The Federal Trade Commission has warned tech companies that changing privacy policies to retroactively scrape old data could be "unfair or deceptive."

3. How are some companies trying to balance their need for data with user privacy concerns?

• Google stated that it only allows a small test group of users, with their permission, to have their personal emails used to train its AI.
• Snap provided "upfront notices" to users about how it uses data to train its AI chatbot "My AI."
• Meta has said its AI did not read messages sent between friends and family on apps like Messenger and WhatsApp unless a user tagged its AI chatbot in a message.