Privacy in Personal AI

Part 1 of 4 in the series "The Road to Privacy in Personal AI".

• Part 1: Privacy in Personal AI

• Part 2: Three Core Challenges in Data Encryption

• Part 3: Solving the holy grail of “Encryption in Use”

• Part 4: Building Privacy into KIN, and Personal AI

In a world where Artificial Intelligence (AI) seamlessly intertwines with our daily lives, and our individual thoughts the call for absolute privacy rings louder than ever. 

In this long-form article, we’ll delve into the background of digital privacy, why privacy matters, data encryption, data trust, data storage, confidential computation, proving information without sharing data, and how we’re approaching this with Kin.

Everyone will own a personal AI

Just imagine having 10 years’ worth of personal data, conversations, tracking etc., and someone “turns off your friend”, as Elon Musk phrased it. It’s a devastating risk.

Personal AIs will be your most valuable digital asset in the future. It cannot be kept in walled gardens and it must transcend ecosystems. It must ultimately be owned and controlled by you - its user.

At Kin, we’re committed to a future where interactions between you and your personal AI are fully confidential. We believe it will be a prerequisite to building the foundation of trust necessary to fully leverage AI on a personal level.

The road to full privacy is a multi-step journey, that can take different turns, but all lead to the same destination; fully confidential interactions between you - the user - and an AI.

Why privacy will matter more than ever

“Privacy Is Dead And Most People Really Don’t Care” as Neil Sahota headlined his Forbes article from 2020. 

At a glance, this headline seems to resonate with the common sentiment. Ask around, and you'll likely hear, “I’m not concerned about privacy, they already have all my data, so it’s okay”. Indeed, the terms and conditions of platforms like Instagram or TikTok rarely cause a stir among their vast user bases.

Yet, when we delve a little deeper, a different narrative unfolds. It’s not merely about digital privacy, but, as Jamie Smith articulated in his LinkedIn post from September 2023, it's more about digital intimacy.

Try imagining walking into your office and seeing your colleague standing with your phone in his hands, browsing through your messages.

How would you feel about that?

Now, even though this is more difficult, try to imagine engaging with ChatGPT-5 or similar, as your Personal AI, which has now become an integral part of your daily routine. You decide to explore its now super-advanced “Custom Instructions” feature to personalize your interaction further. Suddenly, it proactively poses some deeply personal questions:

1. What are the three most important qualities you seek in a romantic partner?

2. You mentioned your son has autism. How many friends does he have, and has he ever acted aggressively towards other children during a meltdown?

3. What's a lie you've perpetuated over the years, one that might have shaped your identity?

4. Have you ever led a double life or cheated in a relationship?

Extremely personal and intimate. Doesn’t feel safe to share. 

Picture an AI having access to this information. What if, at some point, someone in your life, or someone with influence over your life, gains access to this data? The implications are profound. 

The necessity for personal AIs to access such information lies in the evolving user-AI relationship. Over time, the line between the user and AI blurs. The AI becomes an extension of self, much like how we view our smartphones today. For a truly personalized experience that empowers you to optimize your life by leveraging AI’s capabilities, a level of digital intimacy is inevitable. Holding back on sharing personal information could hinder the effectiveness of personal AIs. They thrive on data - your data, which is a reflection of your behaviors, memories, and ultimately, your identity. Without your memory, your identity holds no value.

As we march into the digital future, our dependency on digital identities will only intensify. A decade from now, the digital documentation of our lives will have grown exponentially. With everyone having a personal AI, envision ten years of accumulated personal information residing in this AI.

The digital future presents three paths: escape (avoid using it), defend (rely on regulations), or attack (demand control, ownership, and full privacy).

To fully unlock the potential of personal AI, intimacy is essential. This intimacy hinges on comprehensive trust. Achieving such trust necessitates complete privacy, which in turn, demands ownership and control over your data and your personal AI.

Let’s now get more technical with Part 2.

Part 2: Three Core Challenges in Data Encryption