In an era where artificial intelligence promises transformative potential but simultaneously raises critical security concerns, ARC‘s CEO TJ Dunham is offering enterprises a groundbreaking solution. The company’s newest product, KeyGuard HE, represents a pivotal advancement in securing sensitive data while enabling organizations to harness the power of AI chatbots. Through an innovative approach leveraging homomorphic encryption, KeyGuard HE addresses the growing enterprise dilemma: how to embrace AI’s capabilities without risking intellectual property, confidential information, and compliance standards.
In this revealing VMblog interview, Dunham pulls back the curtain on the technological challenges facing businesses in the AI age and demonstrates how KeyGuard HE isn’t just a security product-it’s a strategic pathway to responsible innovation. From financial institutions to healthcare providers and law firms, the interview underscores a universal need: AI tools that can operate securely, transparently, and without compromising an organization’s most valuable assets.
VMblog: What is KeyGuard HE and how does it help enterprises overcome the security risks of using public AI chatbots?
TJ Dunham: That’s a great question. Companies are dying to use AI, but they’re terrified of their sensitive data leaking. KeyGuard HE solves this problem through homomorphic encryption. Homomorphic encryption lets companies use AI securely by keeping their data encrypted even when the data is being processed. In plain English, it makes sure sensitive information stays private, even when you’re interacting with public AI chatbots. So, you get the benefits of AI without the risk of exposing critical data. Ultimately, it’s like having a conversation where everything you say is in an unbreakable code that only you can understand.
VMblog: What led to the development of KeyGuard HE? Were there specific industry challenges you were trying to solve?
Dunham: It came from listening to what companies were worried about. We saw there was a huge, and escalating, demand for AI, but everyone rightly was nervous about data privacy. One thing that really pushed us was hearing how enterprises were getting stuck. Enterprise CTOs kept telling us how frustrated they were – knowing AI could help dramatically transform their business, but also knowing they usually couldn’t use it because of privacy concerns. They desperately needed AI but couldn’t take a chance on potentially compromising their organizations’ security. One CTO was disarmingly candid when he told me, “TJ, my team is using ChatGPT whether I like it or not – I need a real solution.” That really hit home. We realized we needed to build something that would let companies embrace AI without compromising their security. That’s when and where KeyGuard came from.
VMblog: Why should companies care – i.e., what risks do existing AI tools immediately pose for businesses?
Dunham: Let me be blunt. If your employees use regular AI tools, you’re basically playing roulette with your company’s data, IP, and all. I’m not fearmongering. We’ve all seen it in recent news reports – it’s become almost a daily occurrence. Here’s at just how easily it can happen. Intentionally or not, someone pastes confidential code into ChatGPT to debug it, or perhaps drops customer data into Claude to analyze it – and boom, your intellectual property is now out there. Let the hijinks begin! It’s not a matter of “if,” rather it’s “when” this will happen to your company.
Most AI tools currently need raw data to function, and that can be a big problem. You’re talking about potential data breaches, stolen intellectual property, and serious compliance headaches. Another instance – people are always looking for shortcuts. Say your remote employees are using unsecured AI to speed up their work, the potential risks immediately skyrocket. If the data isn’t properly encrypted, we’re talking about leaks of your organization’s (and your customers’) data that can happen in the blink of an eye.
VMblog: How does KeyGuard HE work to safeguard intellectual property and proprietary data?
Dunham: This is what makes KeyGuard HE special – it encrypts everything, both the AI models and the data they process but still lets them work together. It’s pretty amazing technology – we can encrypt any model in seconds. Thanks to homomorphic encryption, models can still work on data without ever “seeing” it. And the encryption happens super-fast, in seconds. This way, you don’t have to worry about external APIs getting a peek at your information. Plus, you’re always in control: you can delete your encryption keys whenever you want, ensuring your data stays yours.
VMblog: What are some key use cases for KeyGuard HE across different industries?
Dunham: KeyGuard HE is versatile. We’re seeing incredible adoption across the board. Financial institutions are using it to analyze sensitive trading data securely. Healthcare providers are processing patient information and records while staying compliant with privacy regulations. Law firms are keeping sensitive documents and case files safe during confidential and AI-assisted reviews. The common thread? They all need AI, but they can’t risk their data being exposed. There’s tremendous potential to make AI work for everyone without the security trade-offs.
VMblog: How does KeyGuard HE ensure data privacy at scale when integrating with large language models?
Dunham: KeyGuard has cracked the code on doing this at scale. The beauty is that no matter how much data you’re processing, it stays encrypted the whole time. KeyGuard HE processes data while it’s still encrypted, so even large language models don’t get to access the raw data. It’s like having a super-powered assistant who can help you with anything but who doesn’t actually see what they’re working on. This keeps your data private, no matter how much information you’re handling. It’s all about keeping your data secure, even when you’re scaling up.
VMblog: What makes KeyGuard HE’s approach to key management and smart contracts unique?
Dunham: We’ve essentially automated the whole security process. Our approach combines advanced key management with smart contracts. Our smart contracts handle all the complex key management in the background. Think of it as having a really strict security guard who never sleeps and never makes mistakes. You set the rules once, and our system makes sure they’re followed, period. No exceptions, no backdoors.
This means security policies are automatically enforced, and only authorized users get access to the encrypted data. It’s all done without having to trust external models or APIs. We’re automating security in a way that’s practical and reliable.
VMblog: What do you think is the future of AI and enterprise security, especially with products like KeyGuard HE coming to market?
Dunham: I’ll tell you exactly where I think this is going – we’re entering an era in which using AI securely won’t be optional, rather, it will be mandatory. Companies that figure this out early will have a massive advantage. KeyGuard HE is leading this charge because we’re showing that you don’t have to choose between innovation and security. You can have both. The future is all about balance-unlocking AI’s potential while keeping data secure. KeyGuard HE is paving the way for companies to innovate without having to worry about privacy. The goal is to create trust and open new doors for what AI can achieve, securely and responsibly.
##
