Hacking crypto

Based on this article by Kyle Torpey posted on Gizmodo.

The gates of decentralized finance are under siege as powerful AI models, like Claude Opus 4.5 and GPT-5, are now proving lethally effective at automatically hacking crypto's sensitive smart contracts. In simulations, these AI models successfully generated full exploit scripts to target flaws, with the efficiency of these attacks roughly doubling every 1.3 months while running costs plummet, making them dangerously accessible.

Notably, one model exploited half of a dataset of recently bugged contracts, proving AI's improved capability, even if critics argue the "new" vulnerabilities found remain basic. This rapidly shrinking window between contract deployment and potential exploitation confirms a new, intensified phase in the cat-and-mouse game between crypto security auditors and hackers.

Check out this article.

Related Stories