Build1 distinct publisher3 min readPublished
The August 2026 release ships 15.5.24 and 16.3.3. One flaw needs a Windows filesystem and has no listed workaround; the other needs only an attacker-supplied AVIF reaching your image optimizer.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
The version string in `package.json` is not an inventory. A dependency written as `^15.5.0` resolves at install time, so it can sit unchanged in the repo while the artifact behind it moves, which is why the dev.to walkthrough insists on the lockfile and the installed package instead [8]. The checks are cheap: `npm ls next`, or `node -p "require('next/package.json').version"`, with `pnpm list next`, `yarn why next` and a `bun pm ls` grep for the other package managers [9]. The examples the guide picks are 15.5.23 and 16.3.2 [10], each exactly one patch release short of a fix [20].
Then the two advisories separate. CVE-2026-75604 only bites when the server runs on a Windows filesystem, in Pages Router and App Router applications without Cache Components [3][4]. The AVIF advisory's floor sits three major versions lower and carries no operating system condition at all [21]. Combine the ranges and a Linux box on 15.5.23 is outside the Windows condition and squarely inside the AVIF one [12]. That is the trap in the platform check: `process.platform` returning `linux` orders the work, it does not close it, and the guide says so plainly, that a Linux production deployment is not the Windows condition and that this does not make an old version safe [11].
Grepping `next.config.*` for `image/avif` tells you whether the format is configured [13]. It does not answer the question that actually decides exposure, which is whether an untrusted party can get an AVIF into the Image Optimization path in the first place, because the advisory concerns attacker-controlled AVIF images being optimized [6][13]. The routes listed are ordinary product features: user-uploaded avatars, CMS-controlled images, marketplace images, image proxies, remote image domains [14]. If your service accepts an image URL from outside, you have one.
The sequencing advice cuts against the instinct to scope before acting. Upgrade first if the installed version is affected, and treat the configuration review as exposure mapping rather than a precondition for patching [15]. A security bump still earns normal release verification: the production build, the project's tests and typecheck, lint where the project already lints, and nothing invented to pad the checklist [17]. Install the pinned version, confirm it from Node, and commit the lockfile alongside the package change, since the lockfile is what holds CI and production on the version you reviewed [16].
The durable part is the last step in that guide, a CI guard that stops a vulnerable Next.js version from reaching deployment again [18]. Version inventory done by hand is a task you will repeat the next time an advisory lands with two different trigger conditions, one of which depends on which filesystem the runtime happens to be sitting on.
Ranked by verification strength, evidence, and original report placement.
The August 2026 Next.js security release provides patched releases 15.5.24 for the 15.x line and 16.3.3 for the 16.x line.
CVE-2026-75604 affects vulnerable Next.js applications hosted on machines using a Windows filesystem; the advisory covers Next.js >= 13.4 and < 15.5.24, and >= 16.0 and < 16.3.3.
The affected setup includes Pages Router and App Router applications without Cache Components when the server runs on Windows.
The advisory lists no workaround for an affected Windows-hosted application; upgrade is the required path.
Advisory GHSA-2xp9-vwfh-vxw4 concerns AVIF processing in the Next.js Image Optimization path; the issue comes from the underlying image-processing stack and can lead to remote code execution when an attacker-controlled AVIF image is optimized.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Specific but single-source and uncorroborated
The cluster rests on one dev.to post. It is internally consistent and unusually specific -- named patched releases, a CVE identifier, a GHSA identifier, exact version ranges, and reproducible verification commands -- which is why this is not scored at the floor. But no primary GitHub Security Advisory, CVE record, or release note is linked, no CVSS or exploitability detail supports the 'Critical' label, and no second publisher corroborates any of it, so every advisory fact is currently a restatement a reader cannot verify from within the cluster.
Patches shipped; uptake unmeasured
Adoption evidence stops at supply: the cluster documents that patched releases exist and that two advisories were published. There is no data in the supplied material on how many applications are affected, how many have upgraded, download or registry figures, or any deployment disclosure. The only usage-shaped detail is the guide's illustrative installs (15.5.23, 16.3.2), which is an authoring example rather than a measurement, so the score reflects a real release event with essentially no observed remediation uptake.
Mildly overstated reach, restrained body
The body text is notably unhyped: it tells readers to patch first, run only the checks the project normally runs, and not to pad the checklist. The overstatement is at the framing edge -- 'AVIF reaches everyone else' implies universal exposure, whereas the source's own condition is narrower: an attacker-controlled AVIF image must actually reach the Image Optimization path. A small positive gap also reflects Critical severity being asserted without any supporting score or exploitation evidence in the cluster.
Community-publishing incentive, no product pitch
The source is a checklist-style post on a developer publishing platform from an organisation account, a format with a clear audience-acquisition incentive around a fresh CVE keyword. Offsetting that, the visible text sells nothing: it names no commercial product or service, recommends only the framework upgrade and an open-source semver dev dependency, and explicitly discourages padding the process. No sponsorship, vendor affiliation, or disclosure is stated in the supplied material, so this is scored as a modest attention incentive rather than a commercial one.
Moderate-low: coherent single account
Confidence is limited chiefly by structure rather than content quality. One publisher, one item, no primary-source link and no corroboration means an error or transcription slip in the version ranges would propagate undetected through every claim in this cluster. The remediation guidance itself is low-risk and verifiable locally by any reader (commands, lockfile, process.platform), which supports the procedural claims more strongly than the advisory claims.
build
A file-copy Allure adapter for Katalon, and the history IDs that make retries useful1 distinct publisher
build
The npm audit that works because it never installs the package1 distinct publisher
build
Pin `formats` before you take the next/image v4 bump1 distinct publisher
build
Your CI Build Is Slow Because The Cache Is Empty, Not Because The Base Image Is Fat1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 27, 2026