Bringing trust to AI on next generation hardware from NVIDIA and Intel

Problem

As multi-agent AI systems scale across public and private sectors, organizations face mounting challenges around auditability, policy alignment, data sovereignty, and regulatory compliance. Traditional methods lack the cryptographic rigor and performance efficiency required to govern autonomous AI at scale.

Solution

Verifiable Compute creates hardware-rooted trust through Trusted Execution Environments on Intel processors and NVIDIA GPUs, generating cryptographic certificates for every AI operation. These attestations are timestamped and anchored on Hedera Consensus Service, creating an immutable audit trail that enables real-time governance, auditability, and regulatory compliance.

As a new era of autonomous AI agents emerges, we must evolve our trust in AI systems. Verifiable Compute protects and controls AI data, models, and agents with the industry’s most advanced cryptography. It transforms how organizations enforce AI governance, automate auditing, and collaborate to build safer and more valuable AI.

Jonathan Dotan

Founder

Transforming AI Trust through hardware and blockchain integration

EQTY Lab’s Verifiable Compute represents a paradigm shift in how organizations establish trust in artificial intelligence systems. After two years of intensive research with advisors from Stanford and MIT, the solution addresses the fundamental challenge of AI opacity by creating cryptographic proof of every computation performed during AI training and inference. The framework leverages Trusted Execution Environments (TEEs) available on 5th Gen Intel Xeon Processors, extending trust zones through confidential VMs to NVIDIA GPUs, establishing what EQTY Lab calls an “AI notary system” that operates at the silicon level.

At its core, Verifiable Compute addresses the critical gap between AI capability and accountability. As AI systems become more autonomous and make decisions with real-world consequences, organizations need irrefutable proof that these systems operated according to specified parameters and policies. Verifiable Compute creates a verifiable chain of custody for AI operations, establishing not just what decision was made, but proving the exact computational environment, model version, input data characteristics, and policy constraints that governed that decision. This transforms AI from a black box into a transparent, auditable system where every operation can be independently verified.

Verifiable Compute enables new forms of collaboration where organizations can share AI capabilities across trust boundaries while maintaining sovereignty over their data and models. Regulated industries can demonstrate compliance not through retrospective audits, but through continuous, cryptographic proof of adherence to safety protocols and ethical guidelines. For multi-agent systems collaborating across organizational boundaries, the framework provides the foundational trust layer that allows autonomous agents to interact with verifiable credentials, creating an ecosystem where AI systems can be held accountable in ways previously impossible.

Hedera: The enterprise-grade foundation for verifiable AI

The selection of Hedera as the decentralized foundation was strategic and deliberate. Hedera Consensus Service provides the scalability, speed, and enterprise-grade infrastructure essential for recording millions of AI operations in real-time. When AI operations execute within secure enclaves, they generate attestations—cryptographic proofs that are compiled into manifests and immediately anchored to Hedera. Each computation receives an immutable timestamp, establishing when operations occurred relative to any known security breaches. This creates a comprehensive governance framework that can enforce business policies across sovereign clouds and multiple jurisdictions, enabling organizations to prove compliance with regulations like the EU AI Act.

Breakthrough performance with NVIDIA Blackwell

The deployment on NVIDIA’s groundbreaking Blackwell architecture marks a transformative milestone in Verifiable Compute’s evolution. EQTY Lab became among the first organizations to leverage Blackwell’s industry-first TEE-I/O capable GPU, achieving unprecedented 400,000x performance improvements compared to conventional cryptographic methods. This breakthrough enables millions of AI agents to operate with advanced models while maintaining cryptographic verification. Combined with NVIDIA NeMo microservices, the Blackwell integration delivers real-time verification and enforcement of safety tools that govern emergent agent behavior with minimal overhead, fundamentally evolving sovereign AI guarantees to meet the security challenges of autonomous systems.

Summary

Verifiable Compute stands as a testament to the real-world potential of cryptography and web3. With its launch in 2026, it is poised to offer a powerful new chapter in the adoption of enterprise AI and decentralized technologies across industries as varied as finance, life sciences, media, and the public sector, where the requirement for greater trust and transparency is at an all-time high.