Companies Are Sharing Sensitive Data With Frontier AI Labs
Over half of US workers use AI in their jobs, raising concerns about data security. Experts warn companies about relinquishing ownership of proprietary information.
Tech leaders, including Alex Karp and Satya Nadella, have issued warnings that companies may be inadvertently transferring proprietary data to frontier artificial intelligence laboratories. A recent Gallup survey indicates that 52 percent of US workers utilize AI in their professional tasks, commonly for writing, research, and general assistance.
David Steinberg, founder and CEO of Zeta Global, highlighted that the core issue extends beyond employees inputting sensitive information into chatbots. The critical factor is whether the AI architecture preserves a company's ownership and control over its data.
"When a company chooses to build on a shared AI model instead of one intentionally designed to keep its data separate and secure, they're doing more than just sharing data," Steinberg stated. "They're essentially handing over their alpha that sets them apart from their peers, piece by piece."
According to Steinberg, each prompt sent to a frontier AI lab can represent a small portion of a company's data leaving its control. When this occurs across multiple employees and teams, it can amount to a significant leak of a company's most valuable asset.
Machine learning researcher Saugat Nayak agreed, adding that while AI productivity gains are evident, the risks are not solely competitive but also regulatory. Companies need clarity on data residency, retention policies for training, ownership of created intelligence, and vendor transition protocols.