AI and Confidential Computing: A Virtuous Cycle of Value and Innovation

The rise of generative AI is driving huge demand for confidential computing. Confidential computing, in turn, is enabling big new opportunities for technical innovation in AI. These two themes were prominent among industry leaders at the 2nd annual Confidential Computing Summit.

"AI and generative AI are huge vectors of growth—enterprises have to at least experiment with these technologies," said Ion Stoica, UC Berkeley Professor, Executive Chairman of Anyscale, Executive Chairman of Databricks, and Board Member of Opaque Systems. "In order to get value from generative AI, what enterprises have is their data."

However, there is a major roadblock. Though sensitive data typically has more value, regulations are growing, limiting what types of data businesses are allowed to use and how they are allowed to use it.

“The last time I looked, just across the United States, there are 700 bills that want to regulate AI in some shape or form," Stoica said. This growing regulatory burden is holding back enterprises from unlocking their data for AI innovation.

Confidential computing provides a solution. "You'll notice that OpenAI has actually put out a bounty for research projects on confidential computing in the generative AI space. Across the board, I'm hearing from OpenAI, Anthropic, and others that confidential computing is really important to all lines of their business," he explained. “As players like OpenAI and Anthropic increasingly rely on confidential computing for their lines of business, adoption, use, and implementation will grow.”

"Confidential computing will simply become pervasive,” Mark Papermaster, CTO and EVP of Technology and Engineering at AMD, added. “And AI will be the application that drives its broad adoption."

Confidential Computing Unlocks AI Innovation for Businesses

In financial services, regulations like GDPR, CCPA, and GLBA impose strict data privacy, security, and breach notification requirements to protect customer information. In high-tech industries, compliance with regulations such as HIPAA (for health data), GDPR, and industry-specific standards like PCI-DSS (for payment data) is critical to ensure data protection and privacy. But this protection comes at a cost. “The more constraints you have, the more you are going to put limits on innovation," Stoica said.

Confidential computing removes these regulatory hurdles, effectively precluding any constraints.

The vision is to be able to leverage data seamlessly without constraints, as Stoica describes: "I want to do whatever I want today...on the encrypted data. That's the dream."

Cloud providers will have an essential role to play in enabling this scalable vision: "The cloud service providers are incredibly capable at running data centers, securing energy and access to compute chips,” Jason Clinton, Chief Information Security Officer at Anthropic, said. “Why invent the wheel when you can take advantage of that scale?"

For businesses to take advantage of that cloud innovation and to run AI on sensitive data, the essential step is confidential computing.

Unleashing Business Data Drives More Powerful AI

Once data is unlocked through confidential computing, experts see opportunities to create smarter AI that can collaborate securely. Anthropic’s Clinton describes a future filled with sub-agents: “AIs calling other AIs using encrypted channels...dozens or hundreds of AI agents all working together to solve a very complex problem."

But that’s impossible without the right protection in place, as investor Jeremiah Owyang pointed out. "The only way AI agents can be everywhere and be effective is if they tap confidential, proprietary data. They're not doing it yet."

Confidential computing is emerging as the key solution to unleash the full potential of generative AI by removing constraints on data usage while maintaining privacy and security. The opportunity is to drive a new wave of AI innovation that can transform every industry.