Big Tech's AI Race Intensifies with Push for Private Data; Google Buys Airline's Internal Emails

The race for artificial intelligence (AI) supremacy among global tech giants has intensified, shifting focus towards the acquisition of private, non-public data. As differentiating AI models solely on publicly available internet data becomes increasingly difficult, companies are now vying for internal corporate information, including emails, internal communications, and schedules.

This strategic move is fueled by the rise of "AI agents" capable of performing tasks like managing bookings, processing payments, and drafting reports. To excel in these real-world applications, AI models require specific, contextual data that is often not publicly accessible. Google's recent agreement to acquire internal operational data from the now-bankrupt Spirit Airlines for $10 million exemplifies this trend. The deal includes employee emails, Microsoft Teams chat logs, spreadsheets, calendars, and marketing and operational materials, which Google plans to use for AI model training and product development after removing sensitive customer information.

Other major tech companies are also actively pursuing private data. Amazon leverages sales data and consumer patterns accumulated within its e-commerce ecosystem and offers AWS services that allow corporate clients to integrate their internal data with AI. Meta utilizes user data from Facebook and Instagram to enhance its proprietary AI models.

“

AI agents require non-public data, which mostly consists of sensitive materials like corporate secrets and personal information, so there is indeed a concern about personal information leakage during the training process.

The increasing reliance on private data raises significant concerns about privacy and the potential leakage of trade secrets. Emails and internal documents often contain sensitive customer information and confidential business details. As the use of non-public work data expands, the challenge of protecting personal information and proprietary business intelligence becomes more pressing.

Experts warn of the risks involved. Hwang Seok-jin, a professor at Dongguk University's Graduate School of Personal Information Protection, noted that the non-public data required by AI agents often includes sensitive corporate secrets and personal information, posing a risk of data leakage during the training process. He emphasized the need for clearer standards and verification procedures for the collection and processing of private data, along with strengthened accountability for violations.

“

It is necessary to clarify the standards and source verification procedures for the collection and processing of non-public data, and to strengthen the responsibility for violations.