Matthew Lowden, originally from the United Kingdom, brings a background in software engineering and cybersecurity, having led engineering teams at TomTom and Zivver. Elizabeth Schermerhorn combines a technical background with years of consulting experience focused on process optimization and workflow improvement. Having previously worked together at cybersecurity company Zivver, secure data handling has become a natural part of their approach. Through Aetherien, they are building an AI platform specifically designed for organizations operating in highly regulated sectors, including healthcare, government, and financial services.

AI is already inside organizations

For Lowden, the motivation to found Aetherien was not a technological breakthrough, but a shift he witnessed inside organizations. “AI is already being used. Employees are experimenting with all kinds of AI tools, and organizations can no longer prevent that.”

According to him, the real problem is not AI adoption itself but the lack of oversight. “Many of these tools learn from the information users provide. That means sensitive business data can end up somewhere organizations no longer control.”

Schermerhorn sees the same challenge. “Organizations want to innovate while maintaining control. Right now, those two ambitions don’t always go hand in hand.”

Security starts by design

That tension became the foundation of Aetherien. Rather than building another AI application, the founders focused on making existing AI models safe for enterprise use. “In sectors like healthcare, government, and finance, there’s almost no room for mistakes,” says Schermerhorn. “Security cannot be an afterthought. It has to be built into the system from day one.”

Lowden believes many organizations make the same mistake.

“There is enormous pressure to adopt AI quickly. But if you don’t first think about how AI should be used securely, you end up creating new risks instead of solving existing problems.”

EU hosting is not the same as sovereignty

According to the founders, data sovereignty is often misunderstood. “Many organizations say they want an EU-hosted solution,” Schermerhorn explains. “But that only tells you where the data is stored.” She argues that several aspects matter: where data is stored, where it is processed, which legal jurisdiction applies, and ultimately who can actually access the information. “That last point is the most important.”

Lowden adds: “You can host data in Europe, but if a foreign authority can still compel access under its own laws, you don’t truly control your data.” That is why Aetherien takes a different approach. “Only the customer should have access to their own data. That’s the principle our platform is built around.”

A secure layer around existing AI

Rather than developing its own language model, Aetherien builds on existing models such as Anthropic’s Claude. “We use established models as our foundation,” says Lowden. “What we’ve built is a secure layer around them.” That, he says, is where the company differentiates itself. “The model operates inside a controlled environment. Data never leaves that environment unprotected.”

He explains that the platform relies on a cryptographic system requiring two separate keys to decrypt information. “One key is controlled by the customer, the other by us. Without both, the data remains encrypted.”

As a result, neither Aetherien nor its cloud providers can access customer data. “Only the customer decides when data can be decrypted.”

Secure without adding complexity

Although the technology behind the platform is sophisticated, the user experience is intentionally simple. “We designed it to feel like software people already know,” says Schermerhorn. “Users log in with Microsoft, open a familiar chat interface, upload documents, and start working.”

Behind the scenes, strict access controls ensure employees only see information they are already authorized to access.

“We’re not creating new data pathways or exposing additional information.”

Lowden believes usability is essential. “If secure software becomes too complicated, people simply won’t use it. They’ll go back to public AI tools.”

Most AI projects fail for the same reason

According to Lowden, many AI initiatives fail before they ever deliver value. “There is a statistic suggesting around 85 percent of AI projects fail. In many cases, organizations start without a clear objective.”

Schermerhorn agrees. “Sometimes AI is used where traditional automation would solve the problem more efficiently. Not every problem requires AI.”

Lowden believes organizations should resist the hype.

“There’s a lot of pressure to ‘do something with AI.’ But if you don’t understand where it creates value, you’ll end up building solutions nobody actually needs.”

Europe should compete on trust

Lowden acknowledges that Europe’s leading AI models have fallen behind their American competitors. “The gap between Mistral and the leading U.S. models has only widened over the past year.”

For Schermerhorn, however, that does not mean Europe should copy Silicon Valley. “The investment gap is enormous. We’re not going to win by playing the same game.”

Instead, she believes Europe has a different competitive advantage. “Europe can lead in trust, governance, and secure AI.”

Lowden expects those strengths to become increasingly valuable. “Organizations aren’t just looking for powerful AI. They’re looking for AI they can safely deploy.”

AI should accelerate organizations, not create risk

Toward the end of the conversation, both founders return to the same central message. “Companies underestimate how much AI is already being used internally,” says Lowden. “And they underestimate the risks that come with it,” Schermerhorn adds.

For Lowden, the first step is obvious.”Stop putting sensitive business information into public AI tools without proper oversight.”

Ultimately, Aetherien does not define success by scale alone. “Success means customers know their data is protected and that we’re delivering measurable value,” says Schermerhorn. “That means saving time, improving processes, and allowing people to focus on work where human expertise matters most.”

Lowden leaves no doubt about his conclusion: “AI should be an accelerator, not a risk.”