Core Lightning's Silent Emergency: AI-Assisted Attacks Just Forced Bitcoin's L2 Into Lockdown
The message arrived on a Discord channel most node operators check once a week. It was short, technical, and carried the weight of an oncoming storm: restart your Core Lightning node in --offline mode. Not later. Not after the next batch of blocks. Now.
The date was late August 2025. The Core Lightning team, the developers behind one of the three major implementations of the Lightning Network, had found something. Something serious enough to demand that every single node operator disconnect from the network voluntarily. Something serious enough that they refused to disclose what the bug actually was for two weeks.
This is not a routine patch. This is the sound of a system under siege.
Core Lightning, or CLN, is not an obscure side project. It is the backbone of Bitcoin's layer-2 scaling ambitions, written in C by Blockstream, and deployed by advanced users and sophisticated routing nodes across the globe. When its maintainers issue an emergency bulletin, the entire ecosystem should feel the tremors. And this bulletin was issued with an urgency that borders on panic.
The command was deceptively simple: run your node with the --offline flag. In practical terms, this means your node stops routing payments, closes its connections to peers, and waits. But here is the critical nuance that most outsiders miss: the team explicitly advised against shutting the node down entirely. A powered-off node cannot monitor the blockchain. It cannot detect a forced channel closure. It cannot defend the funds locked in its channels. By staying online but disconnected, the node keeps its watchful eye on the chain, ready to respond to any malicious attempt to steal funds. It is a defensive posture, not a retreat.
The fact that this guidance was necessary reveals the nature of the threat. This is not a denial-of-service bug that makes the software crash. This is a vulnerability that could allow an attacker to extract value from channels. This is a capital loss event waiting to happen.
Third-party developers confirmed the severity. Calle, the developer behind the Cashu ecash protocol and a prominent figure in the Bitcoin security community, used the term "severe" without hesitation. The official team messaging was more measured, as corporate communications often are. But the discord between the two tones speaks volumes about the gravity of the situation.
What makes this event particularly unsettling is the timeline. This was not an isolated incident. It was the fourth major Bitcoin infrastructure security alert in four weeks. The sequence reads like a coordinated campaign: Coldcard, the hardware wallet, suffered a vulnerability that resulted in the theft of $114 million in Bitcoin. Boltz, the atomic swap service, announced an indefinite suspension of operations. BTCPay Server, the popular payment processor, issued an urgent update-or-shutdown directive. And now, Core Lightning.
Four attacks. Four critical components of the Bitcoin ecosystem. Four separate teams scrambling to contain the damage. This is not coincidence. This is a pattern.
And at the center of this pattern is a terrifying new development: artificial intelligence is now the primary tool for finding these vulnerabilities.
The Core Lightning team explicitly mentioned that they were verifying "AI-generated CVE reports from multiple sources." This is not theoretical research. This is not a proof-of-concept. This is the real-world deployment of AI to dismantle the security of Bitcoin's infrastructure. The Bitcoin Red Team, led by Calle, reported 85 critical vulnerabilities across 390 projects. That is an astonishing hit rate. The era of the lone hacker manually tracing code is over. The era of the automated hunter has begun.
Let me pause here and share a perspective born from years of auditing smart contracts and protocol implementations. I have spent countless hours manually tracing ERC-20 functions and mapping attack surfaces. It is slow, meticulous work. A single vulnerability can take weeks to find. The idea that an AI model can scan thousands of codebases, identify the subtle logic flaws that lead to fund theft, and generate standardized CVE reports is both impressive and deeply terrifying. The asymmetry is the problem. Attackers can automate their search. Defenders are still largely relying on human review. This is a fundamental mismatch that will only get worse.
The release strategy adopted by the Core Lightning team underscores the severity. They released signed binary files first, with the source code to follow. This is an unusual move. Typically, source code is published first to allow for community review. By prioritizing binaries, the team is signaling that they believe the vulnerability is being actively exploited or will be imminently. They also withdrew support for all previous versions, including the recent 26.04 release. This is a scorched-earth policy. There is no safe version except the one they are about to hand you.
The two-week embargo on vulnerability details is standard practice for responsible disclosure. It gives users time to patch their systems before attackers can weaponize the information. But it also creates a window of uncertainty. For two weeks, the entire Lightning Network will be operating in a state of suspended animation. Nodes will be offline. Routing fees will be lost. And a cloud of suspicion will hang over every transaction.
The economic impact is indirect but real. Node operators rely on routing fees for their income. A two-week offline period is a direct hit to their revenue. Smaller operators, who may be running nodes on a shoestring budget, might decide that the risk is not worth the reward and exit the ecosystem entirely. This could reduce the decentralization of the Lightning Network, concentrating routing power in the hands of larger, better-funded players. That is a slow-moving structural change, but it is a change nonetheless.
Market prices, as is often the case, have been surprisingly resilient. Bitcoin's price is driven by macro factors, not by the technical health of its layer-2 solutions. But the psychological impact cannot be overstated. Institutional investors, who are already cautious about the complexity and risk of crypto infrastructure, are watching these events unfold. Four major security breaches in a month will not inspire confidence. It will reinforce the narrative that Bitcoin is still a frontier technology, and frontiers are dangerous places.
There is also a deeper philosophical question at play here. The narrative around AI and crypto has been largely positive, focused on AI agents transacting on blockchains, autonomous trading bots, and decentralized machine learning markets. This event flips that narrative on its head. AI is not just a user of these systems; it is now a hunter of their weaknesses. The same technology that promises to make markets more efficient is also making them more fragile.
Fragility is the price of infinite composability. I have written that sentence before, and it has never felt more relevant. The Lightning Network is a masterpiece of engineering, a system of channels and hashes that enables near-instantaneous transactions. But its complexity is also its Achilles' heel. Every new feature, every new implementation, every new integration adds another surface for an attacker to probe. And now, the probing is automated.
The response from the Core Lightning team has been professional and competent. They followed the playbook: immediate warning, interim mitigation, and a plan for a patched release. But the speed of the response is not the issue. The issue is the speed of the attacks. AI does not sleep. It does not take weekends off. It does not get tired. The defenders are humans. The attackers are machines. This is a mismatch that will define the next decade of cryptocurrency security.
What happens next is uncertain. The patched version of Core Lightning is expected by the end of September, but the details of the vulnerability will only be released after the two-week embargo. The community will then be able to assess the actual damage. Will there be reports of stolen funds? Will there be stories of nodes that were drained before the warning went out? Or will this be a case of a near miss, a vulnerability discovered before it could be exploited?
The answer to that question will determine the trajectory of Bitcoin's layer-2 ecosystem. If funds were stolen, the trust in Lightning will be severely damaged. If the attack was thwarted, this will be a warning shot, a demonstration of what is possible.
Hype creates noise; protocols create history. The noise around this event will fade in a few weeks. But the history is being written right now, in the form of offline nodes, withheld information, and a growing sense that the machines are learning how to break what we have built. The question is not whether we can fix this vulnerability. The question is whether we can fix the systemic vulnerability that allows a single AI model to threaten the entire network. That is a question that has no easy answer.
Until then, keep your nodes offline. Watch the chain. And remember that in this new world, the most valuable skill is not writing code. It is finding the flaws in code before the machines do.