At 2 a.m., my phone buzzed and spit out a voice message. My brother, gasping, said he’d been scammed by scalpers—someone told him to transfer $400 just to get into the Kansas stadium. The pause and the bitter smirk at the end were all his—except this wasn’t my brother. It was AI synthesized from a few seconds of audio in a social video. Even voiceprints can be forged. Imagine AI carrying out strategies for your vault: a stop-loss instruction gets put on-chain without passing any authorization gate, and you might not even have time to react.

Behind the scam @Suyay shared in the plaza today is the fact that AI voice phishing has surged by 1,200% year over year, and Americans lost $893M to it last year. Most conversations about AI in crypto still revolve around smarter models. Newton Protocol’s question is different: how can we let AI interact with a blockchain autonomously—without relying on black-box trust? Their approach is a secure rollup designed specifically for AI strategies and automated trading, and they use VaultKit to connect rule-checking tools to the vault.

But there is a risk here that’s easy to overlook—just like the caution in the elf-story: Newton doesn’t help you judge whether a strategy itself is good or bad; it only ensures that the pre-set rules have been strictly followed. If there’s a loophole in your take-profit ratio, or you end up importing an incorrect authorized address, Rego will still diligently enforce the wrong policy. The bottleneck of trust isn’t technical execution; it’s whether the path you’ve set in the first place can stand up to scrutiny.

Newton Protocol translates this layer of checks into a machine-readable flow: before a transaction is put on-chain, you must run Rego first (the pre-transaction rule checker). For example, if a withdrawal exceeds 30% of the total vault, or the transfer address is not on the whitelist, or it’s outside a pre-defined time window, Rego will block it outright and not proceed. Only when all conditions are met will it produce an attestation (clearance proof). Then the transaction is packaged by the operator nodes participating in verification. The key here is to hard-code the pre-execution rules into the execution chain—not to assign blame after the fact.

Stop judging by the hype of AI narratives; instead, look at the actual pass/fail records of Rego policy on the Newton Mainnet Beta. The verification logs for these attestations are publicly accessible. How often transactions get rejected—more than the $NEWT price—tells you whether this rule-check system is truly protecting users’ money. @NewtonProtocol $NEWT #Newt