Linus Torvalds Suggests Linux's AI-Driven Patch Flood May Be Easing

Linus Torvalds suggests in his weekly kernel update that the flood of AI-prompted patches, dubbed the "patchpocalypse," may be peaking. The Linux 7.4 rc2 release is one of the largest in kernel history, partly due to code deferred from rc1 and partly due to numerous "various AI fixes." Since Anthropic launched its Mythos bug-finding model in April, AI tools have uncovered far more software flaws than before.

The impact extends beyond Linux: Microsoft patched a record 974 flaws in September, following record fix sprees in June and July, while Oracle addressed 1,449 flaws in a single month. IT teams are struggling to keep pace, especially as attackers increasingly exploit flaws discovered by AI. Torvalds notes that eBPF lead developer Alexei Starovoitov claims fixes are now landing faster than new bug reports on that subsystem, calling it a sign the tide may be turning, though most other areas haven't reached that point yet.

The deluge is taking a toll on maintainers. Lorenzo Stoakes, a principal software engineer at Arm, posted to the kernel mailing list that he is dropping general review duties because AI-generated patch series leave him without the bandwidth to review outside his own maintained areas. He says he simply wants his inbox to be "less horrendous," highlighting the growing human cost of AI-accelerated bug discovery.

Read More at the original source →