Verifiable AI took center stage at DAS London with EigenCloud

At DAS London, Eigen Labs’ J.T. Rose argued that crypto’s subsequent leg gained’t be DeFi however “verifiable AI” — agentic programs whose off-chain work might be confirmed on-chain.

Rose forged Eigen’s stack as a cloud-like trio of providers (knowledge, compute, inference) backed by crypto-economic safety and proofs, letting builders hold public-cloud flexibility whereas including verifiability earlier than funds transfer or state updates land on Ethereum.

Three promising use instances are coming to the fore, in accordance with Rose:

• Autonomous buying and selling brokers that should show they adopted threat guidelines;
• Agent-to-agent (A2A) funds the place work receipts set off settlement and;
• Gaming with attestable outcomes.

“The only most necessary limitation we see for [Artificial Intelligence] founders over the subsequent decade is belief,” Rose mentioned. “And not using a mechanism to make AI verifiable up and down the stack — from inference and benchmarking to coaching and identification — we’ll by no means hit escape velocity for the agentic period.”

The widespread thread is programmability meets provability — “programmability like a cloud, assurances like a blockchain,” as Rose framed it — a response to immediately’s trade-off between tightly constrained on-chain execution and opaque off-chain compute.

David Sneider, co-founder of Lit Protocol, largely shares the vacation spot, specializing in the properties which might instill belief in AI brokers, like “guardrails and insurance policies,” synonyms to advertise right execution.

“[As] with something with safety, there’s a number of elements by way of various kinds of assaults and the way you guard in opposition to them,” Sneider advised Blockworks, arguing that runtime enforcement — not simply after-the-fact proofs — is what turns brokers from demos into reliable infrastructure.

In Lit Protocol’s design, Sneider mentioned, a person may inform an agent, “Ship an electronic mail or purchase a bitcoin,” through which case “it communicates with Lit, which validates in opposition to the insurance policies…and provided that the coverage passes, the key administration community then executes.” By “secret administration community,” Sneider is referring to Lit’s TEE-secured, MPC-based key infrastructure.

Taken collectively, Lit handles the “Can the agent act on my behalf?” query. EigenCloud and Google’s AP2 deal with “Can I belief the results of this offchain computation?”

Or, to simplify the complete belief stack:

• An agent asks: Am I allowed to do that? (Lit)
• Execution proves: Did I do what I mentioned? (EigenCloud/AP2)

Rose outlined a spread of alternate options on the verifiability spectrum, corresponding to TEEs, crypto-economic slashing and zk proofs. Lit can emit success/failure indicators and policy-compliance proofs for every execution, however they’re typically stored inner. The roadmap is to floor these attestations “in privacy-preserving methods into shared registries like ERC-8004 and inter-agent communication protocols like A2A,” so a compliant agent doesn’t must be re-audited in each venue.

Restaked ether has held regular between 2-3 million ETH, largely with EigenLayer | Supply: Blockworks Analysis (Entity labels sourced from Dune’s staking flows desk.)