AI Spotted the Flaw Humans Missed for Years: -50% Wipeout for Crypto Token

Eli Ben-Sasson is one of the creators of Zcash, the cryptocurrency that in 2016 promised real privacy in the blockchain world. Back then, threats meant human adversaries: hackers, governments, criminals. A mathematical vulnerability in a currency built on sophisticated cryptography was exactly the kind of thing that experts in the field should have caught—year after year, review after review.

Except they didn't. For nearly a decade, nobody found it. Automated tests passed. Security audits came back clean. The team's cryptographers slept soundly. Then AI showed up.

Here's the interesting (and terrifying) part: a machine learning system analyzed the code, found the flaw, and exposed it. Not out of malice, not in an organized attack—simply because a neural network trained to recognize patterns can spot connections that the human brain, no matter how expert, struggles to see. Zcash researchers discovered the problem through the same medium that could have revealed it to bad actors.

The damage was immediate and brutal: the token's value crashed 50%, essentially erasing half of the market's trust in the security it had promised in a single session. This wasn't a true hack (thankfully), but the vulnerability's discovery was enough to convince the market that the security guarantee was an illusion.

What makes this story different from typical bug stories is the bigger lesson: machine learning isn't just a tool we humans control to do useful things anymore. In this case, AI did what's supposed to be one of the hardest jobs in cybersecurity—finding flaws that escape professional scrutiny for years—with an efficiency that makes human experts look negligent. Not because of superiority, but because it processes code differently than we do.

Security teams globally are now asking the real question: how many other holes are out there, hidden in code we think is safe, waiting for an AI smart enough to find them before someone with bad intentions does?