Core Lightning nodes are being told to disconnect after a wave of AI-generated bug reports turned out to contain something real.

The usual dynamic is the opposite: maintainers drown in AI noise and learn to skim past it. This time the noise had a critical flaw in it.

The interesting part is not that an AI found a bug. It is that the report still needed a human to reproduce it, judge severity, and decide to warn everyone. AI did not replace that work. It changed what the human spends attention on.

For Lightning that is a good trade. Auditing gets cheaper if machines can surface real leads, and the network becomes harder to attack overall. Agents are not going to replace security people, but they might make the good ones more dangerous to attackers.