Skip to main content
CVE Vulnerability Database
Vulnerability Database/CVE-2026-45046

CVE-2026-45046: Gryph Information Disclosure Vulnerability

CVE-2026-45046 is an information disclosure flaw in Gryph that logs sensitive file content to local databases despite filter protections. This article covers the technical details, affected versions, and mitigation.

Published:

CVE-2026-45046 Overview

CVE-2026-45046 is an information disclosure vulnerability in Gryph, a security layer designed to protect AI coding agents. Versions prior to 0.7.0 log sensitive file-write content into a local SQLite database despite documentation claims that filtering occurs at the default log level. The README states the default log level is minimal, while the actual default is standard. At both standard and full log levels, sensitive content persists in fields such as ContentPreview, OldString, and NewString, bypassing the sensitive file filter contract.

Critical Impact

Local attackers with access to the SQLite log database can read sensitive file content captured during AI coding agent operations, including secrets, credentials, and proprietary source code.

Affected Products

  • Gryph versions prior to 0.7.0
  • AI coding agent integrations relying on Gryph's sensitive file filter
  • Local SQLite logging stores produced by Gryph at standard or full log levels

Discovery Timeline

  • 2026-05-27 - CVE-2026-45046 published to NVD
  • 2026-05-27 - Last updated in NVD database

Technical Details for CVE-2026-45046

Vulnerability Analysis

Gryph provides a security layer that intercepts and logs activity from AI coding agents. The product defines log levels intended to control how much content reaches its local SQLite database. The sensitive file filter is supposed to redact protected file content from these stored records.

The vulnerability is classified under [CWE-212]: Improper Removal of Sensitive Information Before Storage or Transfer. Source code review confirmed that file-write payloads retain ContentPreview, OldString, and NewString fields when the log level is set to standard (the actual default) or full. The documentation incorrectly described the default as minimal, leading operators to assume content was filtered when it was not.

Any party with local read access to the SQLite database can retrieve sensitive file contents that AI coding agents wrote, modified, or previewed. This includes API keys, private keys, configuration secrets, and proprietary code that the sensitive file filter was contracted to protect.

Root Cause

The root cause is a mismatch between documented behavior and implementation. The sensitive file filter does not strip content fields at the standard log level. Combined with the documentation error, operators run Gryph with a less restrictive default than expected and sensitive content accumulates in the local database.

Attack Vector

Exploitation requires local access with low privileges to the host running Gryph. An attacker reads the SQLite database file directly and parses payload fields to recover content the user expected to be filtered. No user interaction is required, and confidentiality impact is high while integrity and availability are unaffected.

No public proof-of-concept code is available. The vulnerability mechanism is described in the GitHub Security Advisory.

Detection Methods for CVE-2026-45046

Indicators of Compromise

  • Presence of populated ContentPreview, OldString, or NewString fields in Gryph's local SQLite database on hosts running versions prior to 0.7.0.
  • Gryph configuration files or runtime arguments showing log level set to standard or full without explicit operator intent.
  • Unauthorized read access events targeting the Gryph SQLite database path on developer workstations or build agents.

Detection Strategies

  • Inventory all hosts where Gryph is installed and identify versions below 0.7.0.
  • Query the Gryph SQLite log database for non-empty content fields associated with files matched by the sensitive file filter.
  • Audit local user accounts and processes with read permissions on the Gryph database file.

Monitoring Recommendations

  • Monitor file access telemetry for reads of the Gryph SQLite database by processes other than Gryph itself.
  • Alert on Gryph configuration changes that lower the log level filter or expand logged fields.
  • Track package versions of Gryph across developer endpoints to confirm patched releases are deployed.

How to Mitigate CVE-2026-45046

Immediate Actions Required

  • Upgrade Gryph to version 0.7.0 or later on every host running an AI coding agent integration.
  • Rotate any secrets, tokens, or credentials that AI coding agents may have written or modified while running vulnerable Gryph versions.
  • Securely delete or quarantine existing Gryph SQLite log databases that were generated by versions prior to 0.7.0.

Patch Information

The maintainers fixed CVE-2026-45046 in Gryph 0.7.0. Details are published in the GitHub Security Advisory GHSA-f3jg-756w-gm35. The patch aligns the sensitive file filter behavior with the documented log level contracts.

Workarounds

  • Set the Gryph log level explicitly to minimal until the upgrade to 0.7.0 is complete.
  • Restrict filesystem permissions on the Gryph SQLite database to the owning user account only.
  • Store the Gryph database on an encrypted volume to limit exposure if the host is compromised.
bash
# Configuration example: restrict log level and database permissions
gryph config set log.level minimal
chmod 600 ~/.gryph/gryph.db
chown $USER:$USER ~/.gryph/gryph.db

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Default Legacy - Prefooter | Experience the World’s Most Advanced Cybersecurity Platform

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.