However, two of the biggest events weren’t the kind of smart contract feats that AI could engineer.

In one, a North Korean-linked group siphoned about $285 million from the Drift Protocol after a six-month social engineering campaign that gained it admin access. For another, an attacker exploited a single authentication flaw that led to the withdrawal of approximately $292 million from CalpDAO.

Another example came on Tuesday, when Humanity Protocol, a decentralized human identity service, suffered a loss of more than $30 million from a private key compromise. CoinDesk found that a hacker gained access to three out of six private keys on an employee’s laptop.

Therein lies the problem. While the most obvious smart contract prompts may be exactly what Anthropic’s filters are designed to catch, the biggest vulnerabilities don’t require a contract bug.

Ledgers’ Guillem notes that these exploits come from familiar vulnerabilities: social engineering, poor signing flows, exposed keys and human error.

A model like Fable doesn’t need to surrender to full exploitation to change the economics of an attack. It can read public repositories, compare older versions of software, summarize audit reports and draft persuasive messages that detect small human operational errors.

“These exploits are rooted in social engineering and human error.”

A guard, in such an environment, has to protect every key path, every dependency, every signature flow and every privileged account. As AI speeds up the scouting phase, the final signing phase becomes more important. Private keys need to sit somewhere a compromised laptop can’t reach, and users need a reliable screen that shows what they’re actually approving.