Recomputed on every load

Who covers what in cyber security

By Nick Forshteyn · CISO AI
Figures as at 18 November 2026

We read and tag every cyber security story we can find. Across 2936 of them from 34 outlets, the interesting result is not what gets covered. It is how differently each subject is covered: across its 34 stories, dark web is 73.5% one outlet, while extortion, at 52 stories, is split across 13 outlets with none holding more than 15.4%.

How this is measured. Every story CISO AI holds, published between 20 August 2026 and 18 November 2026: 2936 in total from 34 outlets, counting the 51 publishers whose stories arrive through the financial wire as one. 6 outlets silent for more than a week and far longer than their usual gap are marked as last seen and cannot carry a headline claim. Each is summarised and tagged by topic when it arrives, and a story usually carries several tags, so topic shares sum past 100%. Counts are of stories, not words. These figures are computed when you load the page, so they move as the archive grows. The limitations are set out at the end, and they matter.

Some subjects have an owner. Others have nobody.

For each topic we count how many distinct outlets covered it and what share the leading outlet holds. Topics with fewer than 8 stories are excluded, because below that one extra story swings the percentage wildly. The two topics named above clear a higher bar again — 25 stories and 3 outlets — because a headline claim should not turn on a sample small enough for one outlet's back catalogue to move.

TopicStoriesOutletsLeading outletIts share
AI security 424 29 The Hacker News 16.5%
data breach 296 16 The Record 20.6%
Patch Management 277 24 The Hacker News 45.1%
Vulnerability 267 20 The Hacker News 46.1%
vulnerability management 211 23 The Hacker News 24.2%
Phishing 206 23 The Hacker News 28.6%
cryptocurrency 163 10 Financial press wire 71.8%
supply chain 158 24 The Hacker News 20.3%
Stories per topic, for the 8 largest topics. A story usually carries several tags, so topics overlap and the counts deliberately do not sum to 2936.

dark web is effectively SOCRadar. Of its 34 stories, that one outlet wrote 25, and only 7 outlets covered the subject at all. Skip that publication and you miss 73.5% of it.

extortion is the opposite. At 52 stories it is spread across 13 outlets, with the largest, Infosecurity Magazine, holding just 15.4%. No single publication is a reliable way to follow it.

Concentration is not the same as volume

The 7 largest outlets still publishing, The Hacker News, Dark Reading, The Register, The Record, Infosecurity Magazine, Financial press wire, CyberScoop, produced 2102 of the 2936 stories, or 71.6% of everything. That sounds like a concentrated market, and by raw output it is. The topic table shows why that is misleading: the outlets that dominate the total are not the ones that own every subject.

The Hacker News: 662 stories (22.5%) Dark Reading: 286 stories (9.7%) The Register: 265 stories (9%) The Record: 264 stories (9%) Infosecurity Magazine: 248 stories (8.4%) Financial press wire: 235 stories (8%) CyberScoop: 142 stories (4.8%) 27 other outlets: 834 stories (28.4%) 71.6% top 7 outlets
Share of all 2936 stories, by outlet. Outlets that have stopped publishing are counted under "other". Recomputed on every load, so the split moves as the archive grows.

The practical version: a reader following one masthead is not getting a smaller version of the whole picture. They are getting a picture with specific, predictable holes in it, and which holes depends entirely on which masthead.

What this means if you are trying to keep up

If you follow a single source, choose it for what you need to see rather than for general reputation. The Hacker News carries 16.5% of AI security; The Record carries 20.6% of data breach; The Hacker News carries 45.1% of Patch Management.

If you would rather not maintain a reading list, that is what the weekly briefing is for: every one of these 2936 stories is summarised in plain language, and each topic above has its own page, including AI security, data breach, Patch Management, Vulnerability.

Limitations, which are real

  • These are the outlets we ingest, not the whole press. Topic shares describe this corpus. A different feed list would produce different percentages.
  • Tagging is model-generated. Topics are assigned automatically when each story is summarised. It is consistent, which is what matters for comparing topics against each other, but it is not a human taxonomy and it will disagree with one at the margins.
  • Counts are of stories, not significance. A one-paragraph advisory and a three-thousand-word investigation count the same here.
  • Outlets join the corpus over time. A source we started reading recently carries only the stories published since it joined, so its share understates its actual output, and comparisons against long-standing sources favour the incumbents. Shares converge as the archive grows.
  • The window is still filling. Feeds deliver stories dated days earlier, so recent counts rise for a while after the fact. That is also why this page is computed rather than published: a figure written into a page was wrong within hours.
  • Outlets stop. A feed that goes dark keeps its stories in the archive. An outlet silent for more than 7 days and more than 3 times its own usual gap between stories is labelled with the date it was last seen wherever this page names it, and is not allowed to carry a headline claim, but its past stories still count in the totals.
  • The financial wire is one outlet here. It delivers each story under the original publisher's name; counting those as separate outlets inflated the outlet count and made cryptocurrency look more fragmented than the press actually is. Leak-site trackers are excluded altogether: a listing is not a story.
  • Cite it with the date you read it, shown above, since the numbers move as the archive grows.

Comments

No comments yet.

To comment, confirm your email once. We send a sign-in link; no password to remember.

Your name appears with your comment; your email never does. By continuing you accept our terms and privacy policy.