TLDR: OpenZeppelin CEO Manuel Araoz says all DeFi is unsafe as AI coding agents find bugs faster. DeFi has lost over $1.1 billion to hacks in a year.
Key Takeaways:
- DeFi runs on public smart contract code, but defenses were built for human attackers. Hacks have chipped at trust and capital, including cross chain failures.
- Araoz warned that AI coding agents are becoming “superhuman” at vulnerability hunting, citing tools like Anthropic’s restricted Claude Mythos.
- If attackers can scan and weaponize code faster than teams patch, defenders face a hopeless asymmetry, and DeFi TVL may keep sliding.
When the ceiling turns into a speed run, security stops feeling like engineering and starts feeling like damage control. DeFi promised transparency, but AI turns that same visibility into a shopping list.
When the ceiling turns into a speed run, security stops feeling like engineering and starts feeling like damage control. DeFi promised transparency, but AI turns that same visibility into a shopping list.
Q&A
What breaks first when defenders cannot patch every bug quickly enough?
Higher risk concentrates in the fastest moving protocols, forcing conservative deployments like narrower scopes, slower upgrades, and tighter dependency controls.
Why does AI finding vulnerabilities matter more than AI writing exploits in DeFi?
Exploit development still needs targets, but mass discovery reduces the time-to-attack window, making even well reviewed code feel temporary.
How could DeFi teams change processes if AI scanning is happening at machine speed?
They may shift toward smaller contract surfaces, formal verification where feasible, stronger privilege separation, and continuous monitoring with faster release pipelines.
Could the industry move away from fully public code without losing DeFi’s core benefits?
Partial abstraction is possible via off chain computation or guarded upgrades, but the fundamental tradeoffs are composability, auditability, and user trust.
What precedent does this resemble from earlier security eras in crypto?
Each jump in attacker capability has forced baseline standards upward, from manual audits to automated testing to staged rollouts, and now potentially to automated defense and rapid response.
No comments yet. Be the first to share your thoughts!