Bridgewater's 11-Day Exploit Claim Meets CISA's 2026 Catalog Data
Security / analysis
Bridgewater's 11-Day Exploit Claim Meets CISA's 2026 Catalog Data
The catalog confirms more additions and a higher share of new bugs, but it records when CISA listed a flaw rather than when attackers first used it.

Bridgewater Associates told clients on Nov. 13 that the average gap between a vulnerability's public disclosure and its exploitation has fallen from about a month to 11 days. The authors, Hudson Attar and Arham Habib, add that zero-day exploits (flaws attacked before a patch exists) have roughly doubled since the first cyber-capable open-source models appeared.
The public data that can test those claims is CISA's Known Exploited Vulnerabilities (KEV) catalog, which the paper itself cites. The Terminal ran the catalog file (version 2026.11.20, released Nov. 20, 1,790 entries) and counted. The catalog supports a rise in volume and a shift towards new bugs. It cannot confirm 11 days, because it holds no exploitation date.

What Bridgewater actually claims
The paper, titled "Current Efforts Aren't Enough to Slow the AI Cyber Virus," says Bridgewater is managing money on a one-in-three chance of an economically destabilizing cyberattack within the next year. That figure is Bridgewater's own and the text gives no model behind it. The post names defensive programmes at three labs, Anthropic's Glasswing, OpenAI's Daybreak and Google's Fairwind, and describes a surge in potential exploits and CVE submissions without giving counts.
It also compares regulation. Japan allocated ¥10 billion, which the paper puts at about $63 million, to cyber defence. South Korea's October law sets a 24-hour incident-reporting deadline and fines up to 3 percent of sales for repeat incidents. The UK's Cyber Security and Resilience Bill would require a report within 24 hours and a full one within 72. South Korea's privacy law now allows penalties of up to 10 percent of sales for large-scale data leaks, up from 3 percent. The paper treats all of this as evidence that policy trails the threat, and it does not claim any of these measures has yet changed an attack count. For how two of the labs gate their security tools, see the site's comparison of Daybreak and Anthropic's verification programme.
What the 2026 catalog shows
CISA added 186 entries in 2024 and 245 in 2025. Through Nov. 20, 2026 it has added 306, which is 25 percent above all of 2025 with 40 days left.
- Jan17 entries
- Feb28 entries
- Mar26 entries
- Apr31 entries
- May21 entries
- Jun23 entries
- Jul26 entries
- Aug31 entries
- Sept43 entries
- Oct36 entries
- Nov 1 to 2024 entries
Source: CISA KEV catalog file, version 2026.11.20, counted by dateAdded, accessed 2026-11-22
September's 43 is the most of any month since January 2025, when the file shows 14. October's 36 is the next highest of the period. The pace is a quarter higher than 2025, not double.
Newer bugs are a larger share, which fits Bridgewater's direction
A rough proxy for freshness is whether a catalog entry's CVE identifier carries the same year as the day CISA added it. The proxy has a flaw: the year in a CVE ID is when the number was reserved, not when the bug went public. It still moves in the direction Bridgewater describes.
| Period added | Entries | Same-year CVE ID | Share |
|---|---|---|---|
| Jul to Sept 2025 | 51 | 36 | 71% |
| Oct 1 to Nov 20, 2025 | 40 | 28 | 70% |
| Jan to Mar 2026 | 71 | 30 | 42% |
| Apr to June 2026 | 75 | 49 | 65% |
| Jul to Sept 2026 | 100 | 86 | 86% |
| Oct 1 to Nov 20, 2026 | 60 | 51 | 85% |
In the first quarter of 2026, 42 percent of additions were same-year bugs. By the third quarter it was 86 percent, against 71 percent in the same quarter of 2025. That is consistent with attackers reaching new flaws sooner, and equally consistent with CISA clearing a backlog of old entries early in the year.
Where the catalog cannot follow
KEV records the date CISA listed a flaw. It does not record when exploitation began, and it has no zero-day flag, so neither the 11-day average nor the doubled zero-day count can be recomputed from it. The paper's text says Bridgewater built its own incident series from public disclosures, media coverage and corporate filings, adjusted for reporting lags. The charts did not come through in the text retrieved for this story, and the method for the 11 days is not described there.
CISA's own clock has tightened in step. Since Jan. 1, 168 of the 306 entries, 55 percent, carry a due date three days or less after listing, and since Sept. 1 it is 79 of 103, or 77 percent, with a median of three days. The site's earlier count of BOD 26-04 deadlines found 120 of 152 entries on three days, and its November triage comparison shows how the split falls.
Bridgewater is an investment firm selling a view of risk, and its one-in-three figure cannot be checked from outside. What can be checked says volume is up 25 percent and the mix is newer. The next data point is the year-end total. At the October and November pace of 60 additions in 51 days, 2026 would close near 350, and a December well below that would suggest September was a spike rather than a trend.
Sources
More in Security
- 01SignDrop Exploit Turns Microsoft's CLFS Hardening Into a SYSTEM PathThe proof of concept works only on machines where the 2025 log-file authentication mitigation is switched on, and Microsoft has said only that it is investigating.
- 02SAP Patches ICM2SHELL and PROXYGATE, Two CVSS 10.0 Pre-Auth BugsOnapsis found both flaws, reports no exploitation as of Nov. 12, and says the SAP Gateway bug depends on an access-control file most sites never tightened.
- 03GPGME's CVE-2026-129975 Double Free Needs Only a Crafted Key, and 2.3.1 Is the FixGnuPG disclosed five days early because GitLab had to update, and no CVSS score exists yet to rank it.
- 04Dragos Seeded 2 Million Prompts With Canaries; Amazon's Decoys Caught ARTEX Agents Grading Stolen AI KeysTwo reports from November 13 and 19 show AI credentials being harvested and spent by machines, and a Chinese lab fetching secrets from ZDR prompts.