Advanced language models automate reconnaissance, phishing and cross-chain exploits — forcing exchanges to adopt continuous, AI-aware security
A surge in AI-driven attacks has transformed small, state-backed hacking cells into highly efficient crypto theft machines, security experts say. Researchers attribute recent large-scale breaches to attackers who now use advanced models to scan codebases, identify vulnerabilities and replicate exploits across multiple blockchains in minutes.
The strategy has already produced eye-popping results: investigators estimate North Korean-linked groups have stolen roughly $2 billion in crypto so far this year, including an alleged $1.5 billion breach of a major exchange in February. The most notorious unit, the Lazarus Group, is repeatedly linked to the largest attacks.
Kostas Kryptos Chalkias, co-founder and chief cryptographer at a major blockchain firm, warned that AI — not quantum computing — is the immediate threat.
“AI is the best tool I’ve ever had as a white-hat hacker,” Chalkias said. “And you can imagine what happens when it’s in the wrong hands.”
Attack chains are increasingly automated. Large language models handle reconnaissance by scanning open-source smart contracts, flagging mirrored vulnerabilities and suggesting attack code. AI also fuels phishing, deepfake impersonations and automated laundering, allowing small teams to move proceeds through mixers and OTC channels with unprecedented speed and subtlety.
Security firms report that every new model release reveals fresh attack vectors. Chalkias urged defenders to match pace:
“Each new version of GPT or Claude finds different weaknesses. If you’re not testing against them, you’re already behind.”
DeFi platforms are particularly exposed because shared code patterns mean a single flaw can propagate across protocols. Experts now expect regulators to mandate continuous, AI-aware red-teaming for exchanges and smart-contract platforms — live audits that rerun vulnerability scans whenever major AI models update.
While quantum computing remains a long-term concern, Chalkias stressed the urgency of the current crisis: AI accelerates exploitation today and could even hasten future cryptographic risks. “The combination of AI and quantum is what freaks me out,” he said. “But right now, AI alone is breaking things at a rate of knots.”
The takeaway for custodians, wallets and exchanges is clear: embed AI-based defenses, enforce continuous audits, and assume attackers will automate discovery and laundering — because they already are.
Disclaimer
This content is for informational purposes only and does not constitute financial, investment, or legal advice. Cryptocurrency trading involves risk and may result in financial loss.