Wasmtime flaws could let malicious WebAssembly code escape its sandbox
Wasmtime is a tool that runs WebAssembly programs in a sealed box so they cannot touch the rest of a computer. Two serious bugs could let a malicious program break out of that box.
| Severity | CriticalVulnerability |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | YesFixed versions were released; the exact version numbers are not stated in the available material. · since 26 Oct 2026 |
| How sure is this? | 1 source, unconfirmedOnly one publisher so far, details may change |
How the attack works simplified
- 1Malicious authorCrafts a WebAssembly module
- 2Module compiledVulnerable Wasmtime build compiles it
- 3Module runsExecutes inside the intended sandbox
- 4Sandbox escapeFlaw lets code break out of the box
- ResultModule gains access beyond its sandbox
What happened
Two critical vulnerabilities in the Wasmtime WebAssembly runtime could allow a malicious WebAssembly module to escape its sandbox. One of the flaws affects only ARM64 systems.
Are you affected?
You are affected if you run a vulnerable Wasmtime build that compiles untrusted WebAssembly modules, and the ARM64-specific flaw applies only to ARM64 systems.
Vulnerable Wasmtime builds (exact versions not stated in the source).
What to do
- Update Wasmtime to the fixed version released by the project
- Prioritize updates on ARM64 systems, where one of the flaws applies
- Only compile and run WebAssembly modules from trusted sources until patched
Timeline
- 26 Oct 2026, 11:52 UTCPublicly disclosed
- 26 Oct 2026, 11:52 UTCReported by The Hacker News · The Hacker News
- 26 Oct 2026, 13:13 UTCPicked up by Zero Radar
Sources 1 report, oldest first
Security press established security newsrooms
- Two Critical Wasmtime Flaws Could Let Malicious WebAssembly Modules Escape SandboxesThe Hacker News · 26 Oct 2026, 11:52 UTC
Technical details
- ID
- ZR-a5411bb3
- Disclosed
- 26 Oct 2026, 11:52 UTC
- On Zero Radar
- 26 Oct 2026, 13:13 UTC (1 h after disclosure)
- Last change
- 26 Oct 2026, 13:13 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works