Miami's Bitcoin Roots Meet AI-Found Bugs

An AI scan found bugs in Bitcoin Core, and the project went public with them fast. Miami's bitcoin and security players are watching the shift.

The people who guard bitcoin’s code used to keep some bugs quiet for up to a year. AI-found bugs have ended that habit.

Bitcoin Core, the program most bitcoin nodes run, disclosed three vulnerabilities on Wednesday. An AI scan found all three, CoinDesk reported.

The finding matters beyond bitcoin. It shows how cheaply AI can now find flaws in code that secures real money.

What the Scan Turned Up

The most serious flaw was flagged on Oct. 23. A computer connected to a node running Windows, macOS or BSD could flood it with data until it crashed.

Developers confirmed the problem the same day. They shipped fixes on Oct. 27 and published details on Wednesday. Operators who have not updated are the ones still exposed.

A node is the software that checks and relays every bitcoin transaction.

The second bug shows why the project changed course. The scan reported it on Sept. 15, and another developer found the same flaw independently six days later.

That flaw let a connected node claim to run an old protocol version and then stall. Messages piled up in the victim’s memory, according to the advisory.

The third bug was rated low severity. It surfaced three times, once in the scan and twice in separate human reports.

Why Core Changed Its Rules

Until recently, Core developers kept medium-severity flaws secret for as long as a year. The delay gave node operators time to upgrade first.

On Oct. 22 the project changed that for bugs it calls “easily discoverable.” Large language models have made such bugs cheap to find, so secrecy buys little.

Now the fixes ship to every supported version as soon as possible, with a public advisory attached. A human contributor checks each AI finding before it reaches the security team.

Three more medium and high-severity bugs, fixed in version 30.0, will be disclosed in two weeks.

The logic is simple. If a machine can find a bug in days, an attacker’s machine can too.

Fast disclosure has a cost. Operators must patch quickly, and attackers learn where to look. But a quiet fix no longer protects anyone if AI tools can rediscover the flaw.

Miami’s Bitcoin and Security Threads

Miami has its own stake in the theme. Hut 8, the Miami-headquartered company, launched in 2017 as a bitcoin mining business.

MAIN reported that it has since become a power-first platform and closed a $1.07 billion credit facility for AI data centers. The coin that built it now shares a stage with AI.

On the research side, FIU’s CIERTA center researches adversarial AI and LLM security. MAIN covered Miami’s FIU already studying AI swarm attacks.

The bitcoin findings are the defender’s version of that work. AI can attack code, and it can also audit it.

Anthropic raised a related alarm this month. Its researchers say small AI models may be months from behaving like worms across cheap smart devices, as MAIN reported. Outside experts urge caution.

Miami also hosts security builders such as Miami’s doxx.net, which raised $38 million to build agentic networking.

Bitcoin is a useful test case because its code is public and heavily watched. If AI finds bugs there, less scrutinized software has more to fear.

That includes the payment, logistics and health systems Miami companies build. Many have far fewer reviewers than Bitcoin Core.

The scan found three bugs in days. The rest of the industry should assume others are looking too.

Sources: