Recent revelations by privacy expert Kevin Bankston regarding a potential privacy breach by Google Gemini AI have sparked discussion in the tech community.

Bankston wrote that the tool scanned his private PDF files stored on Google Drive without his explicit consent, and attempts to deactivate the feature were unsuccessful.

Google's Gemini has us in its pocket?

Gemini

Google says user data is only used to generate answers to queries, not to train the AI model. Still, the company’s lack of a clear explanation of how Gemini AI works raises concerns. The tool’s own response to Bankston suggests the incident may have been caused by a privacy bug or a failure in Google’s internal systems.

Bankston, a participant in Google’s Workspace Labs program, hypothesized that the problem might stem from his previous involvement in testing new Google products, raising questions about the scope of consent users provide to the use of their data in such programs.

The incident highlights the significant challenges of AI development, particularly in the context of data privacy. As the technology becomes more advanced, the risk of privacy breaches increases. It is therefore crucial that technology companies operate with the utmost transparency and accountability, and that users are fully aware of the potential consequences of sharing their data.

Will AI take away our privacy?

This example should prompt us to exercise more oversight over how we want AI to be developed and used. Should we push for a more stringent legal framework for how tech companies use data? Should we demand more transparency and accountability from them for their actions?

The Strawberry Project is set to take ChatGPT to the next level

These are questions we need to find answers to, and fast. The last few months have seen a growing number of scandals about how our data is often used to train AI models without our explicit consent. I think an optional opt-in, allowing our data to be used for training or not, would be a good place to start.

What do you think? Let me know in the comments!

Source: tom's Hardware
The author of the entry is Krzysztof Żwirski