Bitcoin Core just did something it has never done in fifteen years. It gave up on keeping secrets.
On Thursday, October 22, 2026, the maintainers of the software running most of the Bitcoin network merged a change to their security disclosure policy. The reason was not a new attack, not a fork war, and not quantum computing. The reason was that large language models have made some bugs so cheap to find that a months-long embargo protects nobody.
Read that again. The people who build Bitcoin’s most important software looked at what a chatbot can do and decided the old rules were a liability.
Quick Summary
- Bitcoin Core merged a new vulnerability disclosure policy on October 22, 2026, proposed by Chaincode Labs developer Antoine Poinsot.
- Bugs flagged as “easily discoverable” — including anything a simple LLM prompt can surface — now get backported to every maintained version and shipped in a point release immediately, with a public advisory.
- Old policy: low-severity flaws went public two weeks after the next major release. Medium and high severity sat under embargo until two weeks after the last affected release hit end of life. That could be a year.
- The trigger: a volunteer group called the Bitcoin Red Team ran AI models across 390 bitcoin repositories, logged 4,962 findings in about a day, and rated 85 as critical.
- Poinsot’s own words: “The embargo period achieves little but it delays the availability of a fixed release.”
What Bitcoin Core Actually Changed
The policy is not abstract. Here is the exact mechanic, because the details are where the story lives.
Under the old rules, Bitcoin Core graded every reported vulnerability by severity and timed disclosure to match. Low-severity issues became public two weeks after the next major version shipped. Medium and high severity issues stayed secret until two weeks after the last affected release reached end of life — a window that could stretch to a full year.
The logic was paternalistic but defensible: node operators upgrade slowly, so give them a year to move to patched software before the exploit recipe goes public.
Poinsot’s argument, in a message to the Bitcoin Development Mailing List, is that this tradeoff breaks completely once finding the bug is trivial:
“We recently experienced an increase in the number of reports for security vulnerabilities that can be easily (re-)discovered. This is largely due to LLMs drastically reducing the bar to find certain classes of issues or identify covert fixes.”
And then the line that should make you sit up:
“For these issues that are easy to find, this is the wrong tradeoff. The embargo period achieves little but it delays the availability of a fixed release.”
Under the new policy, Bitcoin Core can flag a vulnerability as “easily discoverable” during triage — defined as issues findable “with minimal effort, such as through a simple LLM prompt.” The label is not limited to AI-discovered bugs, but nobody is pretending what caused the change.
For those flaws, the project now backports a fix to every maintained major version, cuts point releases as fast as possible, and publishes a security advisory at the same time the fixed software ships. No more year-long silence.
Poinsot admitted the cost openly: fixes will get less testing time before they reach users. Critical bugs — the ones that could enable coin theft at the protocol level, inflate the supply, or cause a permanent chain split — stay outside the standard policy and get handled case by case.
There was also a fight worth noting. Developer Antoine Riard pushed for the option to ship a fix with no advisory at all and withhold details case by case. Poinsot declined, writing that alerting users without details “could create more risk unnecessarily” by triggering a rushed migration among users who were never affected.
The Number That Should Bother You
In late July 2026, sixteen volunteer developers formed something called the Bitcoin Red Team. They pointed AI models at 390 bitcoin repositories. It took them about a day.
They logged 4,962 findings. Eighty-five of those were rated critical.
Within days, a BTCPay Server flaw the group had reported privately was exploited on live merchant servers, draining Lightning nodes. In August, Core Lightning developers told operators to upgrade or disconnect after AI-generated vulnerability reports turned out to be accurate. We covered that one here: Core Lightning told the world its nodes are getting robbed and wouldn’t say who.
And this all landed in the shadow of July 30, when attackers drained 594 BTC in 25 minutes through a Coldcard hardware wallet exploit. Losses across four sweeps approached $114 million. No quantum computer was involved. No nation-state. A hardware wallet that thousands of people were told was the safe option.
That is the actual threat model in 2026. Not a lab in Beijing with a million qubits. A language model, a GitHub search, and a wallet that was shipped with a bad assumption.
The Hypocrisy You Are Supposed To Ignore
Bitcoin Twitter has spent this entire year fighting about quantum.
BIP-360 drafts. Panic threads about Satoshi’s coins. Podcast episodes about whether the curve is doomed by 2032. Endless screaming about a threat that requires hardware nobody has demonstrated at the required scale.
Meanwhile a volunteer team found 85 critical issues in the code that exists right now, in one day, using tools anyone can rent for twenty dollars a month.
The quantum debate is comfortable because it is theoretical. You can be loudly right about a future threat forever and never be tested. The LLM problem is uncomfortable because it is here, it already drained Lightning nodes, and the fix requires you to do something boring: update your software.
Let me say the unpopular part plainly. The people loudest about quantum were, in most cases, not wrong about the risk. They were wrong about the ordering. The future threat got all the attention while the present one quietly turned vulnerability triage into a fire drill.
Here Is The Divide That Actually Matters
Follow this through to the end and you arrive at the only question this site has ever cared about.
If you hold bitcoin through an ETF, a brokerage, or an exchange, you will never know this policy change happened. Your custodian patches their infrastructure silently. Your statement shows the same number. You are insulated from the bug and you are also insulated from every decision made about your property.
If you hold your own keys, this policy change is a personal obligation. A backported point release dropped on a Tuesday means you need to read the release notes on Tuesday. A flagged “easily discoverable” advisory means the clock started the moment it was published, and it started for everyone at the same time — including whoever wants your coins.
That is the deal. Self-custody is not a product you buy once. It is a subscription to your own attention. The ETF wrapper will always be smoother, and it will always be someone else’s keys, someone else’s patch schedule, and someone else’s judgment about what you are allowed to do with your own money.
And notice what Bitcoin Core just proved: the network survives because the software gets fixed in public, by people who argue about it on mailing lists, with the tradeoffs stated out loud. No marketing department approved the sentence “the embargo period achieves little.” A developer wrote it because it was true.
Try getting that admission out of the company holding your ETF shares.
What You Actually Do About It
No theory. Four moves, in order.
- Find out what version you are running. Open your node, check the version string. If you cannot remember the last time you updated it, that is your answer.
- Subscribe to release notifications. Watch the Bitcoin Core repository releases page. When a point release appears with a security advisory attached, treat it as urgent, not as a weekend task.
- Read the advisory instead of the headline. The new policy publishes details alongside the fix. That means you get information at the same time as attackers, which is exactly why speed matters now.
- Keep your seed offline and your node boring. A node you update is a node that protects you. A hardware wallet with a firmware bug is a liability with a nice box. We wrote about how to choose a wallet without getting rekt, and the same principle applies to the software you run.
If you want to stop renting exposure and start owning the thing, that is what this site is for. Bull Bitcoin lets you buy bitcoin and withdraw it to your own wallet in the same order — no custody theatre, no “we’ll hold it for your convenience” speech.
Use coupon code LOVEISBITCOIN at https://loveisbitcoin.com/bull.
FAQ
Does this mean Bitcoin was hacked?
No. Bitcoin Core found no evidence of a network compromise here. This is a change to how the project reports and patches vulnerabilities in its own software. The protocol rules did not change.
Is AI actually finding real Bitcoin bugs, or is this hype?
Real enough that Core Lightning told operators to upgrade or disconnect in August after AI-generated reports proved accurate, and real enough that a privately reported BTCPay Server flaw was exploited on live merchant servers within days of being found.
Should I be worried that fixes now ship with less testing?
The developers themselves flagged that tradeoff. Poinsot acknowledged accelerated releases reduce testing time in development branches. The counterargument in the policy is that an easily rediscoverable bug is not actually being protected by a long embargo — it is just being patched late.
What about critical bugs that could steal coins?
Those stay outside the standard policy. Bugs that could enable protocol-level coin theft, inflate the supply, or permanently split the chain are handled case by case, which means you should not expect a same-day public advisory for them.
Does this affect me if I hold bitcoin on an exchange?
Not directly, and that is the point. Your custodian handles it and you find out nothing. You also have no way to verify they patched anything at all.
Final Thoughts
The most honest thing Bitcoin Core did this week was admit that secrecy stopped working. Not because the developers got sloppy, but because the cost of finding a bug collapsed to the price of a prompt.
Every institution that sells you bitcoin exposure depends on you never thinking about that. They depend on you believing security is a service you buy rather than a discipline you practise. They depend on you treating an app icon as ownership.
The people who run nodes already know the truth: the software gets fixed because someone reads the advisory and clicks update. That is the whole moat. Not a marketing slogan. Not a conference keynote. Attention, applied to your own keys, on a Tuesday.
So here is my question, and I want the honest answer in the comments: when was the last time you actually updated your node — and did you even know which version you were running before you read this?
Sources: Bitcoin Magazine, “Bitcoin Core Ends Long Embargoes for Bugs AI Can Rediscover” (Oct 23, 2026); Decrypt, “Bitcoin Core Rewrites Its Security Policy for a World Where AI Finds Bugs in One Prompt” (Oct 23, 2026); Bitcoin.com News, “Bitcoin Core Changes Security Policy as AI Makes Vulnerabilities Easier to Find” (Oct 23, 2026).
A CHAT PROMPT CAN NOW FIND A BITCOIN BUG IN ONE TRY. BITCOIN CORE JUST ADMITTED IT CANNOT KEEP THEM SECRET ANYMORE.
