AI customers increasingly worry frontier labs could use their data to compete Palantir CEO amplifies private sector's concerns

As companies race to adopt artificial intelligence, a growing concern is emerging among corporate customers: could the very AI firms they rely on eventually become their competitors? The issue was brought into sharp focus at the Lion Forum venture capital conference in the US state of Massachusetts, where Syed Mohiuddin, Anthropic's head of healthcare, was asked a question that many executives are increasingly considering.

"Why should we trust that you won't steal our business?" the question sounded, as the Semafor outlet, whose correspondent attended the event, reports.

"It's a fair question," Mohiuddin replied. "Because we are both a frontier lab that's building models and a product company that has applications."

The exchange reflects growing unease over the dual role played by leading AI developers. Companies such as Anthropic, OpenAI and Google DeepMind not only build the underlying AI models used by businesses across industries, but increasingly develop software products that can compete with existing enterprise services.

Yet the concern extends beyond software. Many AI companies deploy engineers to work directly with customers, helping integrate models into business operations. While intended as technical support, critics argue these teams inevitably gain detailed knowledge of how banks, manufacturers, consulting firms, retailers and other businesses operate.

The issue was also raised by Palantir chief executive Alex Karp during a recent CNBC interview, where he described frustration among business leaders over the industry's direction.

"These people are livid," Karp said. "They're like, 'I am paying for tokens that create no value… and they're going to get my IP.'"

The anxiety has grown as frontier AI labs expand beyond providing foundational models into developing specialised applications.

Anthropic has already introduced products aimed at legal and design professionals—markets traditionally served by dedicated software providers. That has prompted broader questions about what might happen if AI companies eventually move beyond software tools into offering the underlying professional services themselves.

According to an analysis by Semafor, one factor that could limit that expansion is the rapid rise of open-source AI models.

Many of these models, including several developed in China as well as emerging US alternatives such as Reflection, are narrowing the performance gap with proprietary systems while allowing businesses to run AI locally without sending sensitive data to external providers.

For companies handling valuable intellectual property, that distinction is becoming increasingly important.

"They have to use AI labs' products to stay in the race, but doing so requires sending all of their IP, and that is a very uncomfortable thing to do with somebody who might be trying to replace you," Will Wilson, chief executive of software testing startup Antithesis, told Semafor.

Anthropic has rejected the idea that its goal is to displace the businesses building on its technology.

"Claude Code didn't kill Replit and Cursor," Mohiuddin said last month, referring to AI-assisted coding platforms that continue to grow alongside Anthropic's own developer tools.

"We're not trying to be kingmakers and we're not trying to be market replacers. What we're trying to do is elevate the floor of what's possible."

Whether that balance can be maintained remains an open question.

As Semafor noted, the central issue is not simply whether AI laboratories have the technical ability to compete directly with their customers, but whether doing so would ultimately prove more profitable than supplying the technology that enables those customers to succeed.

The answer could shape the future relationship between AI developers and the businesses increasingly dependent on their models.

By Nazrin Sadigova