Google’s Family of GenAI Apps Raises Concerns Over Data Privacy
In a recent support document, Google has shed light on the data collection and retention practices of its Gemini chatbot apps. The document reveals that Google saves conversations with Gemini for up to three years by default, along with related data such as user location, languages used, and devices used. Despite these conversations being disconnected from Google Accounts, human annotators employed by Google read, label, and process the conversations to improve the service. It remains unclear whether these annotators are Google employees or outsourced workers.
To address privacy concerns, Google allows users some control over the retention of Gemini-related data. Switching off Gemini Apps Activity in the My Activity dashboard prevents future conversations from being saved for review, and individual prompts and conversations can be deleted from the Gemini Apps Activity screen. However, Google notes that Gemini conversations, even when Apps Activity is off, will still be saved to a Google Account for up to 72 hours to ensure the safety and security of the apps.
Google advises users not to enter confidential information or any data they wouldn’t want a reviewer or Google to see and use for improving their products and services. While Google’s data collection and retention policies are not significantly different from those of its competitors, the issue highlights the challenge of balancing privacy concerns with the development of GenAI models that rely on user data for self-improvement.
Similar to Google, OpenAI retains all chats with its ChatGPT for 30 days, while other tech giants like Microsoft and Amazon also offer GenAI products for enterprises that do not retain data. Companies are becoming increasingly wary of the privacy risks associated with GenAI tools, with 63% of organizations imposing limitations on the data entered into such tools, and 27% outright banning them. A survey conducted by Cisco revealed that 45% of employees have entered problematic data into GenAI tools, including confidential employee information and non-public files about their employers.
As GenAI tools continue to gain popularity, concerns over data privacy are prompting regulators to take notice. OpenAI faced scrutiny from the Federal Trade Commission, which requested detailed information on how the company vets and protects consumer data used to train its GenAI models. Italy’s data privacy regulator, the Italian Data Protection Authority, also criticized OpenAI for lacking a legal basis for mass data collection and storage to train its models.
The delicate balance between privacy and AI model development is a timely issue that tech companies must address. As consumers navigate this landscape, it is important for organizations to prioritize data security and transparency to build trust and protect user privacy.
In conclusion, while Google’s data retention policies align with industry standards, the concerns raised should be taken seriously. Striking the right balance between data-driven AI advancements and safeguarding user privacy will be crucial for the future of AI technology.