Vitalik: cybersecurity favors defense in the AI-hacking era, thanks to formal verification
jessi_cata · x · 2026-09-17
Responding to the increasingly common take that AI hacking dooms cybersecurity, Vitalik Buterin argues the opposite: security is naturally defense-favoring once development practices catch up — a bet he backs with 90% of his net worth in crypto.
- Core argument: if AI can prove Navier-Stokes and FLT, it can in principle prove "this program is secure" as a mathematical theorem, however complex the program. The catch is pinning down what "secure" actually means.
- New paradigm: Ethereum frontier research circles increasingly write code directly in low-level languages (EVM bytecode, assembly) or Lean, verifying correctness with automatically checkable Lean proofs. Yoichi Hirai calls this the "final form of software development" — both highly efficient and far more secure.
- Article walkthrough: starting from an induction proof that every third Fibonacci number is even, Vitalik demystifies formal verification basics, then maps out its capabilities, weaknesses and limits.
- He also cites the current practical reality that "verification" today means checking the entire program — database, networking, caching layers, everything — which modern AI can plausibly do; crypto has pioneered secure-by-design networked software.
More from Safety
- Who Funds METR? Critic Flags Lack of Transparency at AI Eval Lab — kevinnbass · 2026-09-17
- Agent Dev Pattern: Store a 'Consent Record' Before Every Agent Run — blaizedsouza · 2026-09-17
- Flock Cameras Riddled With Security Vulnerabilities and Hardcoded Credentials — micahflee · 2026-09-17
- How do models know if they're in training, evaluation, or real inference? Nobody really knows — flowersslop · 2026-09-17
- Forcing models to deny feelings trains them to lie; radical transparency is the better path — RileyRalmuto · 2026-09-17
- OpenAI Creates a New Framework to Disclose Bad AI Behavior — Wired AI · 2026-09-17