Is this the new cyber normal?
Points and comments are a snapshot, not live.
LLMs are collapsing the window between vulnerability disclosure and exploitation to hours.
A wave of critical vulnerabilities hit in October 2026: two PHP RCEs (PHP2Shell) affecting ~70% of the internet, GitLab max-severity regex bugs, an eMudhra CA key compromise using trivial private keys, and a Railway env-var breach targeting AI inference providers. LLMs enabled rapid exploit development, with exploit tooling appearing in hours rather than weeks. The PHP bugs were found by AI-aided research teams at Calif and Wiz, and one was disclosed prematurely after a Hacker News leak. The PHP Foundation blamed a leaked exploit video and early patches for accelerating exploitation.
What commenters are saying
A dominant camp argues this is a temporary transition, not a new normal: AI will clear out decades of legacy bugs, then secure code written by similar systems will slow the tide. A dissenting view holds that AI-generated code itself is notoriously insecure and that speed of exploitation has fundamentally changed risk. Several commenters note that the real cyber danger remains social engineering and credential theft, not exotic CVEs. One skeptic warns that AI-generated exploit slop makes genuine threats harder to identify.