Security readout for executives and security teams
Plain-English summary
Keras safe_mode can be bypassed when a crafted .keras model causes TFSMLayer to load an external TensorFlow SavedModel. If a victim loads and runs an attacker-supplied model, code may execute with the victim process privileges.
Executive priority
Treat as high priority for teams accepting external ML models. The business risk is unauthorized code execution inside ML workloads, but exploitation depends on model ingestion behavior.
Technical view
CVE-2026-1462 is a CWE-502 deserialization flaw in keras TFSMLayer version 3.13.0. from_config() unconditionally loads serialized external SavedModel paths without sufficient validation, undermining safe_mode=True and allowing attacker-controlled code execution during inference.
Likely exposure
Highest exposure is in ML platforms, notebooks, CI pipelines, model registries, or applications that import third-party .keras models or SavedModels. Systems using only internally produced, trusted models have lower practical exposure.
Exploitation context
The CVSS vector requires user interaction: a victim must load a malicious model artifact. KEV is false, and the provided sources do not establish active exploitation in the wild.
Researcher notes
The bundle names Keras 3.13.0 and includes an upstream commit, but no fixed release number is provided. Validate fixes against official Keras or vendor packaging rather than assuming all later versions are patched.
Mitigation direction
- Do not load untrusted .keras models or SavedModels.
- Upgrade to a Keras release or vendor package containing the referenced fix.
- Apply relevant Red Hat errata where applicable.
- Restrict model ingestion to trusted, authenticated registries.
- Run model loading and inference in isolated, least-privileged environments.
Validation and detection
- Inventory systems using keras, especially version 3.13.0.
- Identify workflows that deserialize .keras models from external sources.
- Review use of TFSMLayer and TensorFlow SavedModel imports.
- Confirm safe_mode is not the only control for untrusted models.
- Verify applicable upstream commit or vendor errata is present.
Public sources used
Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.
Conservative CVE-to-ATT&CK context
These mappings and lookup hints may be relevant to the vulnerability behavior, CWE, affected product, or exposure path. Glexia-inferred context is not an official MITRE, ATT&CK, CWE, or CVE Program mapping.
ATT&CK lookup starting points
Use these exact CWE pages and searches to review the Glexia ATT&CK library from this CVE's weakness and description context.
CWE-502: Code execution behavior lookup
Code execution and unsafe deserialization weaknesses often justify reviewing execution behavior and process telemetry. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
Open ATT&CK lookupExecution behavior lookup
The CVE wording references code or command execution, so execution technique review may help defensive triage. This is a Glexia inferred lookup path, not an official MITRE, ATT&CK, or CVE Program mapping.
Open ATT&CK lookupCVE-2026-1462 mapping review
Open the CVE-to-ATT&CK bridge for reviewed, inferred, or future official mappings tied to this CVE.
Open ATT&CK lookup- Severity
- High
- CVSS
- 8.8 (3.0)
- Known Exploited
- No
- Published
Vector: CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
CNA and ADP enrichment extracted from CVE v5
These fields come from the CVE record and ADP containers, not from Glexia's Take. They preserve time-varying source decisions such as CISA SSVC, KEV status, CVSS metrics, and provider references.
SSVC decision data
CVSS vector scores
2 official scoresWe collect every scored CVSS vector available in the official CNA and ADP containers. When more than one version is present, the table keeps the source vectors side by side instead of collapsing them into the highest score.
CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H2.85.9@huntr_aiCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H1.85.9redhat-SADPVulnerability scoring details
Base CVSS 3.1 score
7.8HighVector: CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Base CVSS 3.0 score
8.8HighVector: CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Vulnerability timeline
Timeline events are normalized from CVE metadata, CNA source timelines, ADP timelines, and KEV metadata when present.
- CVE reservedCVE Program
The CVE ID was reserved by the assigning CNA.
- ADP timelineredhat-SADP
Made public.
- CVE publishedCVE Program
The CVE record was published.
- ADP timelineredhat-SADP
Reported to Red Hat.
- CVE updatedCVE Program
The CVE record metadata indicates this as the latest update time.
ADP provider summaries
- 2026-04-13T15:02:12.634Z: Reported to Red Hat.
- 2026-04-13T14:55:28.649Z: Made public.
Source materials
- CVE List V5 sourceCVE List V5
- https://huntr.com/bounties/7e78d6f1-6977-4300-b595-e81bdbda331cCVE reference, @huntr_ai
- https://github.com/keras-team/keras/commit/b6773d3decaef1b05d8e794458e148cb362f163fCVE reference, @huntr_ai
- https://access.redhat.com/security/cve/CVE-2026-1462CVE reference, redhat-SADP · vdb-entry, x_refsource_REDHAT
- RHBZ#2457856CVE reference, redhat-SADP · issue-tracking, x_refsource_REDHAT
- https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-1462.jsonCVE reference, redhat-SADP · x_sadp-csaf-vex
- https://access.redhat.com/errata/RHSA-2026:24977CVE reference, redhat-SADP · vendor-advisory, x_refsource_REDHAT
- https://access.redhat.com/errata/RHSA-2026:37275CVE reference, redhat-SADP · vendor-advisory, x_refsource_REDHAT
Products and packages named in the record
CWE details
CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.
Deserialization of Untrusted Data
Deserialization of Untrusted Data represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.