AI-Assisted Hacking Outpacing Traditional Code Reviews, Says NEAR Co-Founder
NEAR Protocol co-founder Illia Polosukhin has warned that AI is making hacking faster and more dangerous, outpacing traditional code review methods.
The Growing Threat
Polosukhin pointed to last October’s AI-assisted hack on Balancer as the starting point of this new era. He explained that relying on “smart people” to manually review code is no longer enough.
With advanced AI models, hackers can now discover vulnerabilities much faster than developers can find and fix them.
The Solution: Formal Verification
Polosukhin strongly recommends shifting to formal verification — a mathematical method that proves code works exactly as intended.
While this approach used to be slow and expensive, AI has dramatically changed that. It can now generate most of the mathematical proofs quickly, which humans then verify for accuracy.
“You don’t need to rely on AI because the AI just produces the proof. You just verify that the proof is correct,” Polosukhin said.
Real-World Example
A recent Zcash project called Project Tachyon successfully used AI-assisted formal verification. It proved that Zcash’s new Ironwood shielded pool is protected against the type of counterfeiting bug found earlier in its Orchard pool.
What once took years can now be done in just weeks thanks to AI.
Broader Impact
Polosukhin believes formal verification should not stop at blockchain smart contracts. He wants it applied to critical internet infrastructure and networking software as well.
NEAR Protocol’s Latest Upgrade
On the same day, NEAR Protocol released a network upgrade. Key features include:
- Support for quantum-resistant account keys
- Automatic network scaling to handle growing activity
