The problem
AI agents don't just assist anymore. They transact, approve, and spend autonomously at machine speed, often with no human in the loop. A vulnerability that steals $1,000 in a human-speed system steals $10 million in a machine-speed system. Not because it's worse, but because it executes thousands of times faster than anyone can respond. If your agent executes 1,000 transactions per second and your team needs 60 seconds to investigate an alert, that's 60,000 potentially fraudulent transactions before you even understand what's happening.

Current guardrails aren't enough

Today's AI security relies on guardrails, monitoring, and policy enforcement. All three were designed for systems where a human is watching. Agentic commerce has no one watching. LLM-based safety judges inherit the same vulnerabilities as the AI they're checking. If an attacker can trick the agent, they can trick the guardrail too. And monitoring is reactive. By the time a dashboard flags unusual activity, thousands more compromised transactions have already gone through.

Why math, not monitoring

Visa doesn't trust that a card is legitimate. It cryptographically verifies it. Banks don't hope wire amounts weren't tampered with. They prove it mathematically. AI commerce needs the same standard. Cryptography turns trust problems into math problems. Math doesn't care about prompt injection, social engineering, or sophisticated attacks. The proof is valid or it isn't. No confidence scores. No "88% blocked." The shift is from observation to verification. From watching to knowing.

Built for Production

• Jolt-Atlas: Production-ready zero-knowledge machine learning framework that enables fast, practical AI verification. Check it out here.
• Verifiable compute should run everywhere: Our proof system runs fast in browsers, IoT devices, and other constrained environments to meet the needs of rapidly evolving agents.
• Developer-first: Train models in PyTorch, TensorFlow, or any ML framework — export to ONNX and generate proofs with JOLT-Atlas. No circuit expertise required.

Use cases

• Guardrails: Human defined policies, enforced with math at 99% accuracy.
• Verifiable AI inference: Prove a specific model ran on specific inputs with a specific output — critical for compliance in regulated industries.
• Multi-agent workflows: Agent-to-agent handoffs with cryptographic receipts. Every query/response in the chain carries a proof.
• Privacy-preserving compliance: Prove policy adherence without revealing proprietary models or sensitive data. Selective disclosure for audit and regulatory requirements.

Founders: Wyatt Benno, Houman Shadab.