The Ongoing Concerns Over Data Privacy and Security in Generative AI

Written by Brandon Lwowski, July 2024

Data privacy and security issues continue to surround Large Language Models (LLMs) and Large Multimodal Models (LMMs). Many people remain cautious about Generative AI, especially when it comes to using copyrighted and personal data. The worries are similar to those of DuckDuckGo users, who value their privacy and don’t want their data shared or used without permission. 

Recent events highlight these concerns. For instance, Google’s Gemini AI has made headlines regarding data privacy. Kevin Bankson, Senior Advisor on AI Governance, recently took to social media to share his concerns over an automatically generated AI summary in a private tax return. He accused the Gemini AI platform of scanning Google Drive files without user consent. Bankson detailed his struggle to disable this functionality, shedding light on a significant privacy issue. 

But the problem doesn’t end with Gemini. AI companies often keep their training data sources behind lock and key. An investigation by Proof News, reported by Wired, revealed that some of the world’s largest AI companies have used material from thousands of YouTube videos to train their models. They did this despite YouTube’s rules against scraping content without user consent. 

Figma, a popular design app, faced its own controversy. Their Generative AI product, designed to help with software development workflows, was quickly disabled due to plagiarism issues. When asked to create a weather app, the AI produced designs almost identical to Apple’s weather app, highlighting the risks of unintentional copying. 

As these incidents accumulate, the public’s trust in these technologies continues to erode. We recognize the power and potential of these models, but ongoing misuse of data without consent might lead to increased regulation. Such regulations could slow down innovation, regardless of the production benefits it can bring to many industries. It’s clear that while Generative AI holds incredible promise, the industry must address these privacy and security concerns to maintain public trust and ensure sustainable progress.