🧯 The Staff Safety Desk

Monday, October 26, 2026

6 stories

Generated with AI from public sources. Verify before relying on for decisions.

🎧 Listen to this briefing or subscribe as a podcast →

Our technical coverage today opens with a second wave of emergency CPython patches responding to the eMudhra certificate compromise. From there, we detail the newly published mechanics behind the Brotli decoder vulnerability, critical execution vectors in Pillow image handling, and strict new evidence gates for automated code reviewers.

Django & Python Ecosystem

Python Issues Emergency Security Releases 3.15.2 Down to 3.11.18 for Compromised eMudhra CA

Following the initial emergency 3.15.1, 3.14.9, and 3.13.17 CPython releases we tracked last week for compromised eMudhra certificates, the Python core team rolled out a second wave of updates on Monday across five active versions (3.15.2, 3.14.10, 3.13.18, 3.12.16, and 3.11.18). The patches continue to harden `_ssl` against forged chains with weak elliptic-curve keys, while also resolving CVE-2026-115557, a tarfile symlink flaw (CVE-2026-114583), and a cookiejar domain-check bug. Maintainers note that because pip vendors its own CA bundle, pip must be upgraded alongside the Python runtime.

Updating the CPython runtime alone leaves background HTTP client calls vulnerable to MITM attacks if your virtual environment's vendored pip CA bundle remains unpatched.

Verified across 2 sources: Linux Compatible · heise online

Brotli Decoder Out-of-Bounds Write Vulnerability (CVE-2026-111497) Exposes Runtimes

Following up on the critical Brotli decoder heap overflow (CVE-2026-111497) we tracked over the weekend, newly published research details the out-of-bounds write mechanics in Google's reference C decoder (c/dec/decode.c): unchecked return values during block-switch handling (CWE-252 leading to CWE-787). The vulnerability, dubbed HTTP2Shell or Node2Shell, extends beyond the Python, Node.js, and Envoy impacts we previously noted, also exposing Bun and .NET environments to memory corruption via crafted compressed streams. The flaw threatens both inbound request payload processing and outbound HTTP fetches triggered by webhooks or background workers.

WAFs inspecting plain text cannot stop this memory corruption flaw because it triggers during low-level payload decompression before application middleware runs.

Verified across 3 sources: Phoenix Security · Wiz Blog · Aikido

GitHub Actions & Supply Chain

Critical Pillow Vulnerabilities Flagged Across Django Image Upload Handling Paths

A security audit report from Sunday reveals that pip-audit flags over 30 CVEs against Pillow 11.3.0, including three critical heap overflow and execution flaws: CVE-2026-112900 (Ghostscript RCE via EPS decoding), CVE-2026-112393 (TIFF write heap overflow), and CVE-2026-112403 (JPEG SOF3 heap overflow). Applications running standard Django ImageField model validation expose these execution vectors directly to untrusted user uploads, requiring an immediate upgrade to Pillow ~= 13.0.0.

Standard Django ImageField validation calls Pillow image verification routines directly, allowing malformed image uploads to trigger arbitrary code execution before file storage policies apply.

Verified across 1 sources: GitHub

Semgrep Build Blocked by Vulnerable MCP and OpenTelemetry Dependencies in Requirements

A protected OSV scanning run on Monday flagged unsuppressed vulnerabilities in Semgrep's requirement lockfiles, blocking pull request merges. The finding highlights mcp version 1.29.0 (GHSA-84m7-p3x7-pcfv for unbounded session retention) and opentelemetry-api version 1.37.0 (GHSA-xcq7-p3pf-p88h for quadratic header parsing). The repository policy explicitly rejects lockfile metadata overrides or manual inline suppressions, requiring verified upstream package updates and clean pip-compile re-runs with full hashes.

When static analysis toolchains pin vulnerable transitive dependencies internally, manual lockfile overrides undermine CI security gates.

Verified across 1 sources: GitHub

AI Slop & Review Patterns

MergeJudge Introduces Evidence-Gated Security Reviews for GitHub Pull Requests

Open-source PR review engine MergeJudge released a Security review mode on Saturday designed to eliminate hallucinated block directives on pull requests. The tool runs deterministic static AST checks first for secrets and dropped security guards, followed by rubric analysis where the AI is strictly forced to quote an exact line from the diff. If the cited line cannot be verified against the pull request diff, the finding is flagged for human inspection rather than blocking the build.

Forcing automated PR review tools to cryptographically verify diff line quotes prevents unverified assumptions about out-of-scope code from blocking CI pipelines.

Verified across 1 sources: DEV Community

Web App Security Literacy

Code-Level Resource-Exhaustion Gate Patched in Resend Webhook Integration

A pull request submitted Monday for the CBTA platform remediates a denial-of-service vulnerability in its Resend webhook receiver. Previously, the endpoint called request.text() to read the full body into memory prior to validating Svix signature headers, allowing unauthenticated callers to exhaust server RAM. The update immediately rejects requests missing signature headers, caps body stream length to return a 413 error on payload overflow, and preserves exact byte streams for signature validation.

Parsing or reading full HTTP request bodies before validating HMAC signature headers allows attackers to crash web servers via memory exhaustion without knowing the webhook secret.

Verified across 1 sources: GitHub


The Big Picture

Low-Level Native C Extensions Expose High-Level Runtimes to Silent Exploitation Vulnerabilities in native C dependencies like Brotli (CVE-2026-111497) and Pillow (CVE-2026-112900) bypass standard web application boundaries. Because these libraries process unauthenticated streams in request bodies or media uploads, high-level Python code cannot rely on standard try/except blocks to prevent memory corruption and remote code execution.

Dual-Layer CA Remediations Highlight Independent Package Bundling Gaps The response to the eMudhra intermediate certificate compromise demonstrates that updating the operating system or base CPython runtime is insufficient. Because package managers like pip vendor their own certificate bundles independently, security updates must be orchestrated across both language runtimes and package tooling simultaneously to eliminate MITM vectors.

Determinism and Evidence Constraints Replace Open-Ended AI Code Review Prompts As open-source maintainers face an influx of low-quality automated PRs, tool developers are moving away from LLM-only review passes. Systems like MergeJudge now require exact diff line quotes matched against deterministic AST checks before flagging vulnerabilities, actively preventing context-free hallucinated blocks.

What to Expect

2026-11-12 — PostgreSQL 14 reaches community End-of-Life (EOL); final security updates scheduled.
2026-11-19 — PostgreSQL 19 general availability release following AI code review defect remediation delays.

Every story, researched.

Every story verified across multiple sources before publication.

🔍

Scanned

Across multiple search engines and news databases

445
📖

Read in full

Every article opened, read, and evaluated

111
⭐

Published today

Ranked by importance and verified across sources

6

— The Staff Safety Desk

🎙 Listen as a podcast

Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.

Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste
Overcast
+ button → Add URL → paste
Pocket Casts
Search bar → paste URL
Castro, AntennaPod, Podcast Addict, Castbox, Podverse, Fountain
Look for Add by URL or paste into search

Spotify isn’t supported yet — it only lists shows from its own directory. Let us know if you need it there.