This article is a deep-dive from JudyAI Lab — an AI engineering playbook series with 100+ published guides, 5,000+ weekly readers across 60+ countries, focused on the practical side of running AI agents, trading systems, and content pipelines in production.

📰 Key Takeaways

Immunefi CEO Amador is warning that the next three to four years will be a make-or-break period for the crypto industry. He says the whole industry will keep facing AI-accelerated infiltration until security teams can put AI’s defensive capabilities to work and build codebases that attackers simply can’t crack. If the industry adopts crowdsourced security more broadly — getting more researchers to turn AI models into defensive tools — this crisis window might shrink to under two years.

Amador’s comments came in response to Anthropic’s release of its latest model, Claude Fable 5, which sparked industry concern that the model could accelerate crypto vulnerability exploitation. Anthropic says Fable 5 has built-in safety mechanisms that automatically route sensitive requests — like anything touching cybersecurity — to a different model, Claude Opus 4.8.

This conversation is unfolding against the backdrop of a string of major hacks hitting the DeFi space. On April 19 this year, hackers drained about 116,500 restaked ETH (rsETH) from Kelp DAO’s LayerZero-powered cross-chain bridge, worth roughly $290–293 million at the time. LayerZero later pointed out that Kelp DAO was running a 1-of-1 Decentralized Verifier Network (DVN) configuration, meaning the entire cross-chain message path relied on a single verifier — a single point of failure — and that LayerZero had actually advised against this setup beforehand. So far, the Kelp DAO hacker has laundered nearly all of the $220 million in stolen funds, and recovery looks unlikely.

💬 JudyAI Lab Take

AI tooling is evolving fast enough that both attackers and defenders are leveling up at the same time, and the Immunefi CEO’s blunt call — that the next three to four years are make-or-break for crypto — is a backdrop every builder shipping on-chain needs to take seriously.

The Kelp DAO incident is a concrete lesson in what bad architecture actually costs. A $290 million cross-chain bridge running entirely on a single verifier — a 1-of-1 DVN setup with an obvious single point of failure — got flagged by LayerZero beforehand and the warning went unheeded. End result: $220 million laundered, funds essentially unrecoverable. For AI builders, the real takeaway here is that once AI drastically lowers the skill floor for attackers, any architecture decision you made as a “we’ll fix it later” stopgap can turn into a very expensive mistake. Amador’s point about crowdsourced security — more researchers turning AI into a defensive tool, potentially compressing the danger window from four years to two — cuts both ways: the defense side needs to be running ahead of attackers at the tooling level too.

If anything in your system leans on a single verifier or a single trust source, now’s the time to go back and question that trust assumption — don’t let a shortcut you took for convenience become the weakest link in the whole system.

📅 Original Article Info

🔗 Further Reading

References