Bitcoin Core Changes Security Policy as AI Makes Vulnerabilities Easier to Find

Bitcoin Core Shifts Strategy as AI Gets Smarter

In an Oct. 22 message to the Bitcoin Development Mailing List, Bitcoin Core developer Antoine Poinsot announced changes to the project’s vulnerability disclosure policy, citing a growing number of security reports involving bugs that can be discovered or rediscovered with minimal effort.

“We recently experienced an increase in the number of reports for security vulnerabilities that can be easily (re-)discovered,” Poinsot wrote. “This is largely due to LLMs drastically reducing the bar to find certain classes of issues or identify covert fixes.”

Large language models, or LLMs, are the technology underlying AI assistants such as ChatGPT. Their increasing ability to analyze software code presents a challenge for developers trying to balance transparency with the need to protect users from attacks.

Historically, Bitcoin Core—the software that underpins the leading cryptocurrency’s decentralized network—has favored lengthy disclosure embargoes, allowing users time to upgrade their software before the public learns about vulnerabilities. The approach reflects Bitcoin’s decentralized nature and a relatively slow culture of upgrading among node operators.

However, those embargoes have also complicated efforts to deliver security patches to users running older versions of Bitcoin Core.

“For these issues that are easy to find, this is the wrong tradeoff,” Poinsot said. “The embargo period achieves little but it delays the availability of a fixed release.”

Under the revised policy, Bitcoin Core developers can now classify vulnerabilities as “easily discoverable” during their initial assessment. The designation includes flaws that can be uncovered with a simple AI prompt, although it is not limited to vulnerabilities found using artificial intelligence.

Rather than withholding fixes until the next major software release, developers will backport patches to every actively maintained major version and release the updates as quickly as possible. A public security advisory will accompany the patched releases, bypassing the usual embargo periods.

The changes were formalized in a GitHub pull request submitted by Poinsot on October 16 and merged on October 22. In addition to creating the new vulnerability classification, the revision modifies language requiring two weeks’ advance notice before public disclosure, clarifying that the requirement applies to vulnerabilities still subject to an embargo.

The accompanying discussion also highlighted potential tradeoffs. Poinsot acknowledged that accelerating security releases could reduce the time available for testing fixes in development branches, while arguing that straightforward, easily verified patches may be preferable to attempts to conceal vulnerabilities that AI tools could uncover anyway.

Other developers debated whether users should sometimes receive warnings before fixes become available, although the adopted policy calls for publishing advisories alongside patched releases.

How the Move Changes Security Disclosures

The changes represent a significant departure from Bitcoin Core’s previous approach, which generally tied vulnerability disclosure to the severity of a flaw and the lifecycle of affected software versions.

Under the existing severity-based framework, low-severity vulnerabilities are publicly disclosed two weeks after a major release containing a fix. Medium- and high-severity vulnerabilities face considerably longer embargoes, with disclosure occurring two weeks after the last affected release reaches end-of-life status, typically about a year after the initial fix.

Those timelines remain in place for vulnerabilities not considered easily discoverable. Critical flaws that threaten Bitcoin’s fundamental security, including bugs capable of creating unauthorized coins or causing permanent network-wide chain splits, continue to require special handling outside the standard policy.

The revision could have immediate implications for users running older Bitcoin Core software, as all actively maintained major versions would be eligible for expedited patches under the new rules.

The policy change does not indicate that Bitcoin’s cryptography has been compromised or that a specific new exploit has been discovered. Instead, it reflects a changing security environment in which developers can no longer assume that withholding vulnerability details will prevent attackers from independently identifying the same weaknesses.

For Bitcoin Core, the emergence of increasingly capable AI tools is shifting the calculation toward making patched software available sooner, rather than relying on secrecy to protect users.

Earlier this year, developers behind the crypto network Zcash said that AI had been used to discover a vulnerability that made it possible to mint unlimited coins, and the network’s privacy-centric shielded pool design made it impossible at the time to prove that it had not been exploited. The price of ZEC plunged at the time due to concerns over potential counterfeit coins.

In July, developers upgraded Zcash to replace the shielded pool with a new, hardened one, and added independent verification to ensure that the ZEC supply had not been inflated. The price of zcash has more than doubled in the time since.