EthereumETH, coin
DevelopingSecurity

Crypto Industry Adopts Two Strategies as AI Improves Bug Discovery

As AI models become better at finding vulnerabilities, crypto projects are split between rapid patching and minimalist code approaches.

Why it matters

AI-driven bug discovery is accelerating exploit timelines and financial losses, forcing fundamental changes in how crypto protocols handle vulnerability disclosure and code architecture.

Key facts

Bitcoin Core merged a security policy change for bugs findable with a simple prompt to an LLM, moving to immediate fixes instead of embargoes up to a year.

Reported by Bankless

The policy change author Antoine Poinsot stated LLMs are drastically reducing the bar to find certain classes of issues.

Reported by Bankless

Two critical PHP flaws found with help from AI models left approximately 1.3 million internet-exposed hosts vulnerable.

1,300,000 hostsReported by Bankless

Initia halted its entire network after an attacker exploited a Day 0 bug to mint tokens and bridge out approximately $2.7 million.

2,700,000 USDReported by Bankless

Initia cofounder Zon described the vulnerability as a Day 0 bug.

Reported by Bankless

There is no confirmation that AI was used in the Initia hack.

Reported by Bankless

DefiLlama data logs 112 hacks exceeding $1 million in 2026 compared to 70 in all of 2025.

112 hacksReported by Bankless

Anthropic research found frontier models' simulated exploit revenue doubling roughly every 1.3 months in 2025.

1.3 monthsReported by Bankless

What happened

Bitcoin Core merged a security policy change for bugs findable with LLM prompts, moving to immediate fixes instead of year-long embargoes. The change's author noted LLMs drastically lower the bar for finding certain issue classes. Separately, Initia halted its network after a Day 0 bug exploit minted tokens and bridged out approximately $2.7 million, though AI involvement remains unconfirmed. DefiLlama data shows 112 hacks exceeding $1 million in 2026 versus 70 in all of 2025, while Anthropic research found simulated exploit revenue doubling roughly every 1.3 months in 2025. The industry is converging on two responses: maintaining agility to patch quickly, or adopting minimalism to reduce attack surface.

How the story developed

  1. Update

    Bitcoin Core merges security policy change for LLM-findable bugs

    Policy change moves vulnerable code to immediate fix and ship instead of up to a year embargo

    Bankless

  2. Update

    Initia network halted after Day 0 bug exploit

    Attacker minted tokens and bridged out approximately $2.7 million

    Bankless

  3. First report

    First report by Bankless

    Crypto's Two Answers to AI Bug Hunters

    Bankless

  4. Update

    Article published analyzing crypto's two responses to AI bug hunters

    Outlines agility and minimalism approaches as industry reorganization

    Bankless

Affected entities

Original sources

PublisherReportRolePublished
Bankless
Crypto media
Crypto's Two Answers to AI Bug HuntersOriginal