LiveActive security incident?Get immediate response
CVE Record

CVE-2026-1462: Safe Mode Bypass in keras-team/keras

A vulnerability in the `TFSMLayer` class of the `keras` package, version 3.13.0, allows attacker-controlled TensorFlow SavedModels to be loaded during deserialization of `.keras` models, even when `safe_mode=True`. This bypasses the security guarantees of `safe_mode` and enables arbitrary attacker-controlled code execution during model inference under the victim's privileges. The issue arises due to the unconditional loading of external SavedModels, serialization of attacker-controlled file paths, and the lack of validation in the `from_config()` method.

HighCVSS 8.8Not KEV-listedUpdated
Discuss exposureBack to CVEs
Glexia's TakeAutomated analysishigh

Security readout for executives and security teams

Plain-English summary

Keras can be tricked into loading attacker-controlled TensorFlow SavedModels from a .keras model even when safe_mode is enabled. If a team opens or runs an untrusted model, code may execute with that user or service's privileges. This mainly matters for AI/ML pipelines, notebooks, and services that ingest external models.

Executive priority

Treat this as high priority for organizations operating AI/ML workloads that consume external models. The business risk is unauthorized code execution in model-processing environments, potentially affecting data confidentiality, model integrity, and service availability.

Technical view

CVE-2026-1462 is a CWE-502 deserialization issue in keras TFSMLayer. Version 3.13.0 is described as affected. from_config() can load external SavedModels using serialized attacker-controlled paths without safe_mode enforcement, enabling code execution during inference. The source bundle includes an upstream commit, but does not name a fixed release.

Likely exposure

Exposure is likely where Keras loads .keras models or TensorFlow SavedModels from users, partners, model hubs, CI jobs, or shared storage. Systems using only internally built and integrity-controlled models have lower exposure.

Exploitation context

No active exploitation is stated in the provided sources, and KEV is false. Exploitation requires the victim to deserialize or run an attacker-controlled model, matching the CVSS user-interaction requirement.

Researcher notes

Key evidence centers on TFSMLayer deserialization, external SavedModel loading, and missing validation in from_config(). Affected version detail is incomplete beyond Keras 3.13.0 in the bundle. Fixed release information is not provided, so remediation should follow vendor guidance.

Mitigation direction

  • Avoid loading untrusted .keras models or SavedModels.
  • Apply vendor updates once a fixed Keras release is identified.
  • Review the upstream commit and Red Hat advisories for remediation details.
  • Sandbox model loading and inference with least-privilege service accounts.
  • Require provenance, integrity checks, and approval for externally sourced models.

Validation and detection

  • Inventory Keras usage across notebooks, services, CI, and model-serving systems.
  • Check whether keras 3.13.0 or affected downstream packages are present.
  • Identify workflows accepting third-party .keras models or SavedModels.
  • Confirm safe_mode is not treated as sufficient protection for untrusted models.
  • Review Red Hat errata applicability for deployed platform packages.
Prepared
Confidence
medium
Sources
9

Public sources used

Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.

Potential ATT&CK relevance

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 · medium confidence lookup

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 lookup
description · low confidence lookup

Execution 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 lookup
cve · low confidence lookup

CVE-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
Vulnerability profileCVE Program record
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

Official CVE source material

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.

2CVSS vectors
5Timeline events
2ADP providers
8Source links

SSVC decision data

CISA-ADPCISA Coordinator
Timestamp
Version
2.0.3
Exploitation: pocAutomatable: noTechnical Impact: total

CVSS vector scores

2 official scores

We 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.

ScoreVersionSeverityVectorExploitImpactSource
8.8CVSS 3.0HighCVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H2.85.9@huntr_ai
7.8CVSS 3.1HighCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H1.85.9redhat-SADP

Vulnerability scoring details

Base CVSS 3.1 score

7.8High
CVSS 3.1 vector shape for CVE-2026-1462Attack VectorAttack ComplexityPrivileges RequiredUser InteractionScopeConfidentiality ImpactIntegrity ImpactAvailability Impact

Vector: CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Attack Vector
NetworkAdjacentLocalPhysical
Attack Complexity
LowHigh
Privileges Required
NoneLowHigh
User Interaction
NoneRequired
Scope
ChangedUnchanged
Confidentiality Impact
HighLowNone
Integrity Impact
HighLowNone
Availability Impact
HighLowNone

Vulnerability timeline

Timeline events are normalized from CVE metadata, CNA source timelines, ADP timelines, and KEV metadata when present.

  1. CVE reservedCVE Program

    The CVE ID was reserved by the assigning CNA.

  2. ADP timelineredhat-SADP

    Made public.

  3. CVE publishedCVE Program

    The CVE record was published.

  4. ADP timelineredhat-SADP

    Reported to Red Hat.

  5. CVE updatedCVE Program

    The CVE record metadata indicates this as the latest update time.

ADP provider summaries

CISA-ADPCISA ADP Vulnrichment
other:ssvc
redhat-SADPkeras: Keras: Arbitrary Code Execution Vulnerability Bypassing Safe Mode
other:Red Hat severity ratingcvssV3_1
  • 2026-04-13T15:02:12.634Z: Reported to Red Hat.
  • 2026-04-13T14:55:28.649Z: Made public.

Source materials

Affected products

Products and packages named in the record

VendorProductVersion / packageStatus
keras-teamkeras-team/kerasunspecifiedListed
Weakness

CWE details

CWE-502

CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.

CWE-502 · source CWE mapping

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.

Open CWE intelligenceSame-CWE CVEs