Amazon Linux 2 and 2023 kernels are out now. EKS and ECS AMIs dated 31 October carry the fix, Redshift follows on 2 November and RDS and Aurora on 3 November.
AWS: Powertools for AWS Lambda (Python) data masking could fail open, and three more AWS libraries need updates
EditorialBy TrustList Editorial
AWS bulletins of 29 September and 1 October 2026: Powertools for AWS Lambda (Python) 3.6.0 to 3.34.0 could return unmasked values on error (fixed in 3.35.0, no workaround), plus fixes for Amazon Ion Python, security-agent-mcp-server and GluonTS.
- Cybersecurity
- Vulnerability Management
- Patch Management
- AWS
- +2 more
About AWS: Powertools for AWS Lambda (Python) data masking could fail open, and three more AWS libraries need updates
AWS: Powertools for AWS Lambda (Python) data masking could fail open, and three more AWS libraries need updates
2 October 2026: Amazon Web Services published four security bulletins for open-source libraries it maintains on 29 September and 1 October 2026. None is a flaw in an AWS service itself; all four are in code that customers install in their own applications, so AWS cannot fix them for you.
Not yet independently verified. This rests on AWS’s own bulletins; no independent report was checked, and AWS gave no severity score for any of them. We read the Powertools bulletin in full and the other three through their summaries. We will update this when it can be confirmed, and remove this note.
Powertools for AWS Lambda (Python): masking that fails open
CVE-2026-104002. Powertools is a widely used toolkit for Lambda functions, and its data-masking utility is meant to hide sensitive fields, such as card numbers or personal data, before they are logged or passed on. In versions 3.6.0 to 3.34.0, an error during masking could return the original value instead of failing. AWS says this "might allow actors to read sensitive field values that the application intended to mask". It is fixed in 3.35.0, where masking errors now raise a DataMaskingError. AWS says "there is no workaround available".
This matters for compliance as much as security: if masked values ended up in logs, those logs may now hold personal or payment data that your records say they do not.
Three more libraries
- Amazon Ion Python before 0.15.0 (CVE-2026-104020): a deeply nested Ion value can exhaust recursion in the reader and crash the application, a denial of service. Fixed in 0.15.0; until then AWS describes disabling the C extension and handling
RecursionError. - security-agent-mcp-server 0.1.1 up to 0.2.0 (CVE-2026-97662): argument injection in the diff scan can create, overwrite or truncate files outside the workspace. Fixed in 0.2.0; meanwhile scan only trusted repositories and run the server with minimal privileges. It is an MCP server that AI coding agents call, so it often runs with a developer's own permissions.
- GluonTS before 0.17.0 (CVE-2026-100308): loading an untrusted model artifact with
Predictor.deserialize()can run operating-system commands. Fixed in 0.17.0; load only artifacts you trust.
What to do
- Search dependency files (requirements, lock files, Lambda layers) for
aws-lambda-powertools,amazon.ion, the security agent MCP server andgluonts, and bump them. - Rebuild and redeploy Lambda layers that bundle Powertools: updating the package locally does not change functions already deployed.
- Review logs written by functions that used data masking on affected versions, and treat any unmasked sensitive data you find as an incident under your data-protection process.
- Treat MCP servers as privileged software: pin versions and review what they can touch.
Sources
- AWS security bulletin: CVE-2026-104002, fail-open error handling in the data masking utility in Powertools for AWS Lambda (Python) - 1 October 2026
- AWS security bulletin 2026-122-aws (Amazon Ion Python, CVE-2026-104020) - 1 October 2026
- AWS security bulletin 2026-121-aws (security-agent-mcp-server, CVE-2026-97662) - 1 October 2026
- AWS security bulletin 2026-119-aws (GluonTS, CVE-2026-100308) - 29 September 2026
Categories & features
- Cybersecurity
- Vulnerability Management
- Patch Management
- AWS
- Data Security
- Software Development
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
You may be interested in
The quarterly advisory covers Java 8 to 27, with a top score of 7.0, and a separate notice adds the 9.1-rated TLS flaw CVE-2026-91403. Amazon Corretto shipped 27.0.1, 25.0.5, 21.0.13, 17.0.21, 11.0.33 and 8u512.
Attackers are encrypting NAS devices through a flaw Synology fixed in December 2024 and points to advisory SA-24:20, whose worst bug, CVE-2024-10441, allows remote code execution without a login.
Kiteworks published 125 security advisories on 30 September 2026, 12 rated critical, for Core, the Email Protection Gateway, Secure Data Forms and MFT Server. The most severe, a gateway code-injection flaw, is fixed in 9.4.1; update to 9.5.1.
IBM's records cover Langflow OSS 1.0.0 to 1.12.8: an authentication bypass and code execution, both 9.8 with no login needed. Version 1.12.8, released hours before the fix, is still affected.
The IDEs ship their own Java runtime and trust store, so an operating system update does not remove the eight compromised intermediates. Fixes run from 2026.2.4 and Toolbox 3.9.1; 2025.1 and older get none.
Faced with a flood of vulnerability reports, the curl project keeps only High and Critical issues private on HackerOne. Everything else goes to public GitHub issues, and old reports will be disclosed.
Every release from 08.01.0102 to 18.00.0004 is affected. An application that builds IN lists from user input can be crashed, or left running on a corrupted heap.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More CybersecurityThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.