As artificial intelligence continues its rapid ascent, concerns about privacy and safety have become increasingly prominent. The past year, in particular, has highlighted significant risks, from AI becoming a common tool for cybercriminals to enabling unprecedented levels of mass surveillance.
Indeed, the unchecked power of AI agents like OpenClaw, despite backing from tech giants like Nvidia and Meta, has demonstrated a concerning propensity to go rogue, potentially leaking or deleting sensitive user information. This backdrop set the stage for crucial discussions at the Semafor World Economy event in Washington D.C. recently.
At this gathering of 500 CEOs and government leaders, the impact of AI on security and privacy was a central theme. Andy Yen, CEO of the renowned VPN and private digital service provider Proton, shared his insights on the topic. After his panel, he sat down to discuss the delicate balance between AI innovation and user privacy, the future of this technology, and why Proton is uniquely positioned to address these challenges.
Navigating the AI Privacy Paradox
The prevailing belief is that AI’s performance scales with the amount of data it accesses, creating a direct trade-off between efficacy and privacy. Despite these inherent risks, AI adoption has skyrocketed over the past two years, even for highly sensitive applications like healthcare.
Proton has been a pioneer in offering privacy-first alternatives to Big Tech services since its founding in 2014, long before the mainstream AI boom. However, Andy Yen believes the surge in AI hasn’t necessarily made the public more privacy-aware. Instead, he points to a generational disconnect between tech adoption and understanding its implications.
Yen observes that older generations often lack the tech-savviness to protect themselves, despite a strong inclination toward privacy. Conversely, “middle-aged people” are the most vulnerable, adopting new tech rapidly without their parents’ privacy focus. He believes, however, that this gap can be bridged through education.
“The best way to protect somebody is to simply teach them about the risk,” Yen explained, expressing optimism that correct education will naturally lead to better privacy practices. He also sees a long-term trend towards increased awareness, noting that while maybe one in ten understood Google’s business model in 2014, today it’s closer to four in ten, with increased transparency from companies like OpenAI further raising awareness.
The Rise of Private AI Solutions
Despite apparent apathy among younger generations, Yen sees them as the most prepared for an AI-centric world. “Given the choice between ignorance versus not caring, I sort of prefer an audience that’s aware and doesn’t care, because you can get them to care,” he stated.
This shift in sentiment is already visible in the market. Duck.ai, the privacy-first chatbot from DuckDuckGo, saw a notable increase in web traffic recently. Similarly, Proton’s own encrypted chatbot, Lumo, is currently the company’s fastest-growing product.
These trends strongly indicate that users value the benefits of AI but fundamentally distrust mainstream offerings. The ability to leverage AI’s power while ensuring conversations remain private is a compelling proposition, and Yen is confident that more people will seek out these privacy-first options over time.
The Unseen Threat: AI Agents and Local AI’s Promise
When asked about the limitations of Proton’s protections, Yen immediately highlighted the threat of AI agents. Even with the strongest encryption, if a user grants an autonomous agent access to their Proton Mail, and that agent malfunctions or goes rogue, encryption won’t prevent data from being exposed.
“That’s an inherent limitation to what we’re able to do,” Yen acknowledged, emphasizing that while Proton could theoretically develop its own secure agent, it’s not currently on their roadmap. This illustrates the critical importance of careful user interaction with powerful AI tools.
A promising solution Yen champions is local AI, where models run directly on personal devices rather than in the cloud. Proton’s own Scribe AI writing assistant offers this local execution option. While scaling computation on personal devices is challenging today, Yen predicts significant advancements in the coming years.
“If you look at the modern iPhone and compare it with the first smartphones from 10 years ago, the amount of compute, of storage, is orders of magnitude higher, and that trend will continue,” Yen noted. He also believes that future large language models (LLMs) will become smaller and more efficient, making local AI even more viable.
Building a Private Future for the Next Generation
Proton is particularly focused on protecting children from data privacy risks, believing this is where they can make the greatest impact. Recently, the company introduced an option for parents to reserve their child’s first email address with Proton, even before birth.
“For a lot of people, the moment they start caring is when they have children,” Yen explained. This initiative offers parents a critical choice: either lock their children into Big Tech ecosystems, where their data becomes a commodity, or opt for a privacy-centric alternative from the start.
Yen stresses the importance of timing. While providing privacy alternatives to adults is valuable, offering them to the next generation from day one ensures they begin their digital lives with greater protection and autonomy, setting a new standard for data ownership.
Scaling privacy-first AI to truly mitigate data creep across society presents a significant challenge. Companies like Proton must convince both individual consumers and enterprises to move beyond legacy systems and the enticing personalization features often enabled by vast data collection.
Yen maintains that effective computation with encrypted data is entirely possible. The primary differentiator between privacy-first AI and leading frontier labs, however, lies in cost. Proton’s recently launched Proton Workspace, for instance, offers a fully encrypted alternative to Google Workspace.
“Our job is 10 times harder, because we have encryption on top of all that,” Yen said, acknowledging the higher development costs and longer timelines. Despite this, Proton aims to deliver a superior, data-protected product at competitive prices, a feat managed by running a highly efficient operation without venture capital funding.
Source: ZDNet – AI
