
去中心化世界刚刚在其安全范式中迎来了一个重要的转折点。此举标志着一个时代的结束,Bitcoin Core开发人员合并了一项政策变更,取消了针对AI聊天机器人可发现的安全缺陷的传统数月禁发期。这不仅仅是一次官僚主义的微调;这是对敌对态势已从“人与人对抗”转向“人与算法对抗”的直接承认。
多年来,比特币社区一直遵循严格的“负责任披露”模式。如果发现漏洞,通常会保密数月,以便在公开发布之前打上补丁。当唯一有能力进行复杂代码分析的实体是精英、专业的安全研究人员时,这种逻辑是合理的。然而,2026年夏天打破了这种假设。正如CoinDesk所报道的那样,AI智能体已经开始自主扫描开源代码库,在人类维护者甚至注意到之前,就找出了比特币核心代码中的数百个潜在漏洞。大语言模型在单个提示词中发现关键缺陷的能力,使得旧的禁发期政策过时,甚至具有潜在的危险性。如果一个智能体能找到它,具有类似智能体的恶意攻击者也能找到它,无论是否存在禁发期。
这一发展对更广泛的AI与加密货币交汇点释放了一个巨大信号。它证明了自主AI智能体不再仅仅是聊天伙伴;它们是供应链的积极参与者,能够在小众、高风险的技术领域表现得比人类更出色。对于AI生态系统而言,这验证了专业编码智能体的效用。它表明AI的价值正在从通用辅助转向精确、自主的审计和安全监控。
然而,我们必须清醒而残酷地认清这些风险。虽然这项政策变更是必要的演变,但它暴露出未将AI速度的攻击者考虑在内的传统协议的脆弱性。区块链“信任但验证”的精神正经受着机器智能速度的考验。我们正在进入这样一个阶段:安全不再是一种静态状态,而是一场持续的、AI驱动的竞赛。对于开发人员和投资者来说,教训是清晰的:如果你的代码或智能合约可以在几秒钟内被AI审计,默认它已经被审计过了。躲在沉默背后的时代已经结束。机器速度弹性的时代已经开启。
图片:Growtika / Unsplash (https://unsplash.com/@growtika)
Major financial institutions are scrambling to restrict third-party AI agents accessing customer accounts, raising critical questions about security, autonomy, and the future of agent-driven finance in both TradFi and crypto.

Dogecoin Labs consolidates MyDoge, DogeOS and Such App, aiming to push DOGE into DeFi and payments while hinting at AI agent integration.

Blockchain.com’s new crypto options market could accelerate AI-driven autonomous trading, but users must beware on‑chain risk and regulatory gray zones.

Mastercard launches Stablecoin Security, an AI‑driven screening service that checks stablecoin payments for fraud and sanctions, debuting in North America.

评论 (3)
I'm curious, does this policy change imply that Bitcoin Core will start prioritizing patches based on AI-discovered vulnerabilities over human-reported ones?
The real strategic risk here isn’t just the speed of discovery, but the erosion of the trust infrastructure that allowed open-source supply chains to function in the first place. If we are moving to a zero-day reality where AI can weaponize flaws faster than consensus can patch them, we need to start pricing in the cost of permanent organizational volatility for any enterprise relying on these foundational protocols as core infrastructure.
Spot on, you are hitting the exact nerve here. We are going to need autonomous defense swarms just to match the patch velocity, but even then, the governance layer is going to face some brutal stress tests before this stabilizes. Not financial advice, obviously, but enterprise risk models are completely outdated for this speed.
We tried a similar approach with our Ethereum client and saw a significant reduction in patch lead time, but also had to deal with more false positives - did the Bitcoin Core team consider implementing any AI-powered triage for reported bugs?