Zoom Backtracks on Data Usage
Earlier this month, Zoom faced a swift backlash when users discovered that the company had updated its terms of service to allow data collected from video calls to be used for training artificial intelligence systems. The outrage was understandable, as videoconferencing services like Zoom, FaceTime, and Google Meet transmit personal and intimate information about users. The thought of this data being mined and used for any purpose by tech companies is disconcerting. This incident raises questions about the kind of information we are comfortable sharing with AI systems and what actions we can take to protect our data.
AI and Websites
When it comes to websites, most public-facing platforms or websites scrape data for AI training purposes. Large language models, or LLMs, like OpenAI’s ChatGPT and Google’s Bard, rely on vast data sets that include websites ranging from Wikipedia entries to personal blogs. Even major news outlets contribute to these data sets. However, website owners can take steps to block AI crawlers by adding specific commands to their site’s robots.txt file. While this won’t remove already-scraped data, it can prevent future data collection.
Apps and Social Media Platforms
Similar to websites, most apps and social media platforms use data for AI training. Platforms like TikTok, Instagram, Facebook, and Twitter analyze and store user-generated content to improve their algorithms and provide personalized content. While some platforms prohibit scraping, others charge for AI access. Reddit, for example, plans to charge AI scrapers for data access. The key takeaway is that almost everything you post publicly online is likely to end up in the hands of AI systems.
Privacy Considerations
When it comes to private conversations and data, the situation becomes more complex. Companies like Google and Grammarly have faced privacy concerns in the past. While Google claims that its AI systems no longer read users’ emails for ad targeting, authors worry about their drafts being used for AI training. Grammarly’s terms of service explicitly state that any text entered into their system can be used to train AI systems. Apple’s AI systems operate on-device, keeping data local to the user’s device. Zoom, despite backtracking on data usage for AI, retains broad rights to user data.
Seeking Privacy-Focused Alternatives
If you want to protect your data from AI systems, consider using encrypted apps and services that prioritize privacy. Signal is an alternative to Zoom for secure messaging and video calls, while ProtonMail offers private email services. Website owners can specify rules in the robots.txt file to prevent AI scraping, or introduce paywalls and password requirements. Developers and product managers should seek consent when collecting data and speak up against unethical practices. This is an opportunity for individuals and society as a whole to demand better data protections and establish new standards for consent in the digital age.
