← Cohort Security

Y Combinator Summer 2026

Decloak scanned 233 companies in this cohort and found a total of 6983 findings across 5 severities.

Security analysis done on 23 August 2026

66

Mean score

63

Median score

233

Members scanned

100%

Coverage (0 failed)

Findings by severity

Top 5 performers

1

Meteoric

meteoric.earth

A
2

Isengard Industries Inc

isengardindustries.com

A
3

Ekpa

goekpa.com

A
4

Praxis Robotics

praxisrobotics.io

A
5

Robocurve

robocurve.org

A

66/100

Mean security score

20%

Graded B or higher

5

Critical findings across the cohort

80%

Affected by "Missing Content-Security-Policy header"

Overview

The Summer 2026 batch of Y Combinator startups presents a mixed picture of web‑application hygiene. Scanning 233 member sites produced an average security rating of 66 / 100, with the bulk (73 %) earning a grade of C. The most striking insight is the prevalence of high‑severity client‑side injection weaknesses: Missing Content‑Security‑Policy headers were found on 186 of the 233 sites, and unsafe DOM‑manipulation patterns (innerHTML assignments and React’s dangerouslySetInnerHTML) appear on a similar scale. While a handful of companies are already demonstrating strong practices, the cohort as a whole shows ample room for improvement in defending against cross‑site scripting and related attacks.

What we found

Critical findings – The scan identified five critical issues across the cohort. Because the data are aggregated for privacy, we cannot break these down by individual startup, but the presence of any critical flaw signals a need for immediate remediation wherever they exist.

High‑severity patterns

  • Missing Content‑Security‑Policy header – Detected on 186 of 233 sites (≈ 80 %). This header is a primary defense against script injection; its absence leaves browsers free to execute any script that can be loaded, dramatically widening the attack surface. The issue was reported 186 times in total, meaning each affected site triggered the check once.
  • innerHTML assignment – Found on 182 of 233 sites (≈ 78 %). Directly inserting HTML strings without sanitisation is a classic XSS vector. The scan recorded 376 total occurrences, indicating many sites contain multiple vulnerable assignments.
  • dangerouslySetInnerHTML usage (React) – Observed on 145 of 233 sites (≈ 62 %). This React‑specific API bypasses the framework’s built‑in escaping, and misuse can introduce the same injection risks as raw innerHTML. It was flagged 441 times, showing that several components per site often employ the API unsafely.
  • Missing Strict‑Transport‑Security (HSTS) header – Detected on 61 of 233 sites (≈ 26 %). Without HSTS, browsers are not forced to use HTTPS for subsequent visits, exposing users to downgrade attacks. Each affected site triggered the check once, for a total of 61 occurrences.

Medium‑severity observations – While the cohort’s high‑severity issues dominate the narrative, a few medium‑level problems also emerged: 156 sites (67 %) lack X‑Frame‑Options, 126 sites (54 %) inject script tags dynamically, and 124 sites (53 %) serve an overly permissive CORS policy (Access-Control-Allow-Origin: *). Additionally, 100 sites (43 %) expose JavaScript source maps publicly, leaking potentially sensitive build information. These findings, though less urgent than the high‑severity injection gaps, still merit attention because they can facilitate information gathering and indirect attacks.

Overall, the tally of findings—5 critical, 260 high, 1 900 medium, 1 275 low, and 3 543 informational—highlights a landscape where many startups have adopted basic HTTPS and informational best practices but still need to harden client‑side defenses.

Who’s leading

Four companies earned an “A” grade in this assessment: Meteoric, Isengard Industries Inc, Ekpa, Praxis Robotics, and Robocurve. Their scores indicate a comprehensive application of security headers, safe DOM handling, and overall robust configuration relative to the cohort. These organisations serve as practical benchmarks for their peers, demonstrating that high‑grade security postures are achievable even within the rapid‑development cycles typical of early‑stage startups.

What this means

The snapshot presented here is diagnostic, not punitive. It shows where the Summer 2026 Y Combinator cohort collectively excels and where systemic gaps remain. Companies that wish to dive deeper into the specific findings affecting their own sites can request the full private report, which includes detailed remediation guidance and prioritized action items.

Improving the highlighted high‑severity issues—particularly CSP adoption and safe DOM manipulation—will raise the overall security grade and reduce the risk of client‑side exploits. As the cohort continues to grow and iterate, regular reassessments will help track progress and encourage a culture of proactive security hygiene across all Y Combinator startups.

Grade distribution

A · 5

B · 41

C · 171

D · 15

F · 1

Most common issues

Ranked by severity, then by prevalence - aggregate across the cohort, never attributed to an individual company. Purely informational checks (e.g. reachability, redirect chains) are excluded.

"Sites" counts each affected company once. The total-detections figure can be higher, since a single site can trigger the same check more than once (e.g. several offending scripts on one page).

highMissing Content-Security-Policy header186/233 sites
highinnerHTML assignment detected182/233 sites (376 total detections)
highdangerouslySetInnerHTML usage detected (React)145/233 sites (441 total detections)
highMissing Strict-Transport-Security (HSTS) header61/233 sites
mediumMissing X-Frame-Options header156/233 sites
mediumDynamically injected script tag detected126/233 sites (413 total detections)
mediumOverly permissive CORS policy (Access-Control-Allow-Origin: *)124/233 sites
mediumpostMessage listener without apparent origin check118/233 sites (140 total detections)
mediumJavaScript source map publicly accessible100/233 sites (758 total detections)
mediumHSTS max-age is too short7/233 sites

Is your company in this cohort?

This page only shows what's safe to publish. Run your own free scan to see your full report - including everything that stays private here.

Claim your report

Methodology & FAQ

What is Cohort Security?

A public, quarterly-refreshed directory of aggregate web security posture across defined groups of companies (accelerator batches, funding classes, industry verticals), built from the same free, passive scan anyone can run against any site at decloak.dev today.

What's published about a named company on this page?

Only observational posture any visitor could already determine for themselves — TLS/HTTPS configuration and security header presence — and only for this cohort's honour-roll top performers.

What about every other company in the cohort?

Every company below the honour roll is folded into the cohort-wide severity and issue-prevalence statistics only. Nothing on this page ever attributes a specific finding to an individual company outside the honour roll, or makes one identifiable by elimination.

How is this data collected?

The same free-tier scan available to anyone at decloak.dev, run once per company in the cohort. No active testing, credential access, or exploitation — see decloak.dev/scanner for exactly what the scanner does and doesn't do.

Can a company request removal from this page?

Yes, no justification required — see decloak.dev/scanner for the removal request form.

More on how this scanner works, and how to request removal →