How bad is it? Tracking the rise in incidents across HN and public data

Tracking outages, vulnerabilities and breaches on the Hacker News front page and other databases

HN is one of my main sources of news for many years now (please never change!), and I can’t be alone with the feeling that the front page recently became full of stories about incidents, breaches, and outages. Now has it always been like this and we’re just paying more attention to these things recently? Or is it really that bad? Let’s find out.

For three years the Hacker News front page carried about 4.6 vulnerability and breach stories a week. Over the last twelve months it was 7.8. In the last four weeks it was 17, almost four times the old normal.

Outage, vulnerability and breach stories on the Hacker News front page

However, this chart could as well be a case of Baader–Meinhof effect (a cognitive bias in which a person notices a specific concept or thing more frequently after recently becoming aware of it)

Public data confirms the increase across the board. The US vulnerability database (NVD) published about 4,200 new CVEs a month in 2025 and 17,481 in October 2026. CISA’s list of vulnerabilities known to be exploited grew by 20 a month last year and by 28 a month this year. Outages are up too, but far less: twenty companies that publish their incident history, among them GitHub, Discord, Vercel and Dropbox, reported 95 incidents a month this year against 80 last year.

New CVEs published · NVD, per month
Vulnerabilities newly known to be exploited · CISA KEV, per month
Incidents on 20 company status pages · GitHub, Discord and 18 others, per month

Why now?

I think we can broadly classify the potential reasons into two categories:

AI on the attacker side. Models now find old bugs and chain small ones into working exploits, so attacks that needed a skilled team are within reach of many more people. Commenters describe probe traffic on their sites quadrupling in two months and incident-response firms turning new customers away. This fits the breach and exploited-vulnerability numbers.

AI on the fixer side. The same models hand defenders hundreds of findings at once, and everyone patches in a hurry. One commenter blames the outages on “hasty fixes to AI-discovered vulnerabilities”: security updates that bundle large changes, shortened test windows, and fixes written by agents and reviewed by nobody. This fits the outages, which are mostly systems breaking on their own rather than being attacked.

Both can be true at once: attackers got faster, and the rushed response is breaking things.