The concentration of AI (artificial intelligence) development in the hands of a few powerful corporations raises significant concerns about individual and societal privacy.

With the ability to capture screenshots, record keystrokes, and monitor users at all times through computer vision, these companies have unprecedented access to our personal lives and sensitive information.

Like it or not, your private data is in the hands of hundreds, if not thousands, of businesses. There are tools on the market that allow anyone to check how many companies have theirs. For most people, it’s several hundred. With the rise of AI, it’s only getting worse.

Companies around the world are implementing OpenAI tech into their software, and everything you enter gets processed by OpenAI’s centralized servers. On top of that, OpenAI’s safety personnel have been leaving the company.

And when you download an app like Facebook, nearly 80% of your data can be collected. That can include things like your habits and hobbies, behavior, sexual orientation, biometric data, and much more.

Why do companies collect all this info?

Simply put, it can be highly lucrative. For example, consider an eCommerce company that wants more sales. If they don’t have detailed data on their customers, they’ll need to rely on broad, untargeted marketing campaigns.

But suppose they have rich data profiles on customers’ demographics, interests, past purchases, and online behavior. In that case, they can use AI to deliver hyper-targeted ads and product recommendations that drive significantly more sales.

As AI weaves its way into every aspect of our lives, from ads and social media to banking and healthcare, the risk of exposing or misusing sensitive information grows. That’s why we need confidential AI.

The data dilemma

Consider the vast amounts of personal data we entrust to tech giants like Google and OpenAI every day. Every search query, every email, every interaction with their AI assistants—it all gets logged and analyzed. Their business model is simple: your data, fed into sophisticated algorithms to target ads, recommend content, and keep you engaged with their platforms.

But what happens when you take this to the extreme? Many of us interact with AI so intimately that it knows our deepest thoughts, fears, and desires. You’ve given it everything about yourself, and now it can simulate your behavior with uncanny accuracy. Tech giants could use this to manipulate you into buying products, voting a certain way, or even acting against your own interests.

This is the danger of centralized AI. When a handful of corporations control the data and the algorithms, they wield immense power over our lives. They can shape our reality without us even realizing it.

A better future for data and AI

The answer to these privacy concerns lies in rethinking the foundational layer of how data is stored and computed. By building systems with inherent security and privacy features from the ground up, we can create a better future for data and AI that respects individual rights and protects sensitive information. One such solution is decentralized, non-logging, private AI powered by confidential virtual machines (VMs). Confidential VMs play a crucial role in ensuring data privacy during AI processing. These VMs are designed to process and store sensitive data securely, using hardware-based trusted execution environments to prevent unauthorized access and data breaches.

Features like secure hardware isolation, encryption in transit and at rest, secure boot processes, and trusted execution environments (TEEs) help maintain the confidentiality and integrity of the data. By leveraging these technologies, businesses can ensure that users’ data remains protected throughout the AI processing pipeline without compromising privacy.

With this approach, you retain full control over your data. You can choose what to share and with whom. Achieving truly private and secure AI is a complex challenge that requires innovative solutions. While decentralized systems hold promise, only a handful of projects are actively working to address this issue. LibertAI, a project to which I contribute, along with initiatives like Morpheus, can explore advanced cryptographic techniques and decentralized architectures to ensure data remains encrypted and under user control throughout the AI processing pipeline. These efforts represent important steps toward realizing the potential of confidential AI.

The potential applications of confidential AI are vast. In healthcare, it could enable large-scale studies on sensitive medical data without compromising patient privacy. Researchers could mine insights from millions of records while ensuring that individual data remains secure.

In finance, confidential AI could help detect fraud and money laundering without exposing personal financial information. Banks could share data and collaborate on AI models without fear of leaks or breaches. And that’s just the start. From personalized education to targeted advertising, confidential AI could unlock a world of possibilities while putting privacy first. In the web3 world, autonomous agents could hold private keys and take actions on the blockchain directly.

Challenges

Of course, realizing the full potential of confidential AI won’t be easy. There are technical challenges to overcome, like ensuring the integrity of encrypted data and preventing leaks during processing.

There are also regulatory hurdles to navigate. Laws around data privacy and AI are still evolving, and companies will need to tread carefully to stay compliant. GDPR in Europe and HIPAA in the US are just two examples of the complex legal landscape.

However, perhaps the biggest challenge is trust. For confidential AI to take off, people need to believe that their data will be truly secure. This will require not just technological solutions but also transparency and clear communication from the companies behind them.

The road ahead

Despite the challenges, the future of confidential AI looks bright. As more and more industries wake up to the importance of data privacy, demand for secure AI solutions will only grow.

Companies that can deliver on the promise of confidential AI will have a major competitive advantage. They’ll be able to tap into vast troves of data that were previously off-limits due to privacy concerns. And they’ll be able to do so with the trust and confidence of their users.

But this isn’t just about business opportunities. It’s about building an AI ecosystem that puts people first. One that respects privacy as a fundamental right, not an afterthought.

As we hurtle towards an increasingly AI-driven future, confidential AI could be the key to unlocking its full potential while keeping our data safe. It’s a win-win we can’t afford to ignore.

Jonathan Schemoul is a technology entrepreneur, CEO of Twentysix Cloud, aleph.im,  and a founding member of LibertAI. He’s a senior blockchain and AI developer specializing in decentralized cloud computing, IoT, financial systems, and scalable decentralized technologies for web3,  gaming, and AI. Jonathan is also an advisor to large French financial institutions and enterprises such as Ubisoft, stewarding and promoting regional innovations.