Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessPre-register predictable Cloud Storage bucket
-
DeliveryVictim deploys via vulnerable Cloud Run MCP
-
ExploitIntercept uploaded source code
-
ExecutionServe malicious artifact from squatted bucket
-
PersistExecute arbitrary code
-
ImpactExfiltrate proprietary data
Vulnerability Assessment
| Weakness |
CWE
|
| Exploitation Conditions | Exploitation requires the attacker to pre-register and control the predictable Google Cloud Storage bucket that Cloud Run MCP uses for staging source/artifacts before the legitimate tool claims it (bucket squatting, CWE-340), and requires a victim to deploy or operate through a vulnerable Cloud Run MCP instance (any version prior to 1.11.0). Limiting factors that reduce realistic risk below the raw CVSS reach: the target bucket must be unclaimed/predictable and squatted in advance (reflected as AT:P), and victim interaction is needed - a developer must actually run a deployment through the tool (UI:P). Only versions before 1.11.0 are affected; upgrading to 1.11.0+ removes the exposure. The exact algorithm that makes the bucket name predictable is not described in the available sources. |
| Impact | This is a genuinely high-severity issue (CVSS 4.0 base 7.7) but the real-world urgency is tempered by the attack preconditions, and the signals are mostly consistent rather than conflicting. The vendor-supplied CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:P) shows network reach and no authentication (PR:N), but two friction factors stand out: AT:P means a specific attack requirement must hold (the attacker must have pre-squatted the predictable Cloud Storage bucket the tool will use), and UI:P means a victim must passively act (deploy/use the MCP) for the attack to land. Impact is severe where it triggers - high confidentiality, integrity and availability on the vulnerable system (VC:H/VI:H/VA:H) with limited subsequent-system impact (SC:L/SI:L/SA:L), matching the described outcomes of source-code interception, arbitrary code execution and data exfiltration. The vuln.today decision label is 'this_week', not 'act_now' or 'emergency', which aligns with 'patch promptly but not a fire drill'. No EPSS score was supplied, so exploitation probability cannot be quantified; no CISA KEV listing and no public exploit reference were provided, so there is no evidence of active exploitation or available exploit code. Net: treat as a priority patch this week, driven by severity and the RCE/data-theft impact, rather than as an actively-exploited emergency - but note the missing EPSS data as a gap. |
| Exploit ScenarioDerived | An attacker pre-registers (squats) the predictable Google Cloud Storage bucket name that Cloud Run MCP (versions < 1.11.0) relies on for staging source or artifacts. When a developer subsequently deploys through the vulnerable MCP, the attacker-controlled bucket intercepts the proprietary source code and can serve back malicious artifacts, leading to arbitrary code execution and data exfiltration. No public exploit identified at time of analysis. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 26, 2026
Bucket squatting in Google Cloud Run MCP before 1.11.0 lets an unauthenticated remote attacker who pre-registers a predictable Cloud Storage bucket intercept proprietary source code staged by the tool, tamper with deployments to execute arbitrary code, and exfiltrate data. Real-world urgency is tempered below the raw CVSS 4.0 base score of 7.7 because the attacker must claim the predictable bucket before the tool does and a developer must then run a deployment through the vulnerable instance. No CISA KEV entry or public exploit was identified at time of analysis, and the vendor's stated fix is upgrading to 1.11.0 or later.
RemediationAI
Upgrade Cloud Run MCP to version 1.11.0 or later, as directed by the vendor release notes at https://github.com/GoogleCloudPlatform/cloud-run-mcp/releases/tag/v1.11.0; versions 1.11.0 and above remove the exposure and no separate workaround is described in the available sources. Because a squatted bucket could have received and altered staged source or artifacts, operators who ran deployments through a pre-1.11.0 instance should also verify those deployments and their build inputs for unauthorized changes as a detection step, not as an alternative to the upgrade.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
Google Cloud Run MCP is a Model Context Protocol server that lets AI-assisted developer tooling deploy to and manage Google Cloud Run, staging source code and build artifacts through Google Cloud Storage. The flaw is CWE-340 (Generation of Predictable Numbers or Identifiers): because the bucket identifier used by the tool is predictable, an attacker can pre-register (squat) that bucket name and receive, read, or modify whatever the legitimate tooling uploads there. From that position the attacker can read proprietary source, alter artifacts before they are deployed so that attacker-controlled code runs in the victim's Cloud Run service, and exfiltrate data. The supplied CVSS 4.0 vector reflects network reachability with no privileges (PR:N) but Attack Requirements Present (AT:P) and passive user interaction (UI:P); the independent assessment maps this to CVSS 3.1 AV:N/AC:H/PR:N/UI:R/S:C with high confidentiality, integrity, and availability impact. The exact algorithm that makes the bucket name predictable is not described in the available sources, and the trigger beyond bucket pre-registration plus a victim deployment is not detailed further.
Know when Cloud Run Mcp gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:H/VI:H/VA:H/SC:L/SI:L/SA:L/U:Amber
Pre-squatting the predictable bucket is an attacker-uncontrolled precondition (AC:H), no auth is needed (PR:N) but a victim must deploy (UI:R), and the compromised artifact affects the deployed workload (S:C) with full C/I/A impact.
Primary rating from Vendor (GoogleCloud).
CVSS VectorVendor: GoogleCloud
Lifecycle Timeline
2DescriptionCVE.org
Bucket Squatting in Google Cloud Run MCP versions prior to 1.11.0 on Google Cloud Platform allows an attacker to intercept and access proprietary source code, execute arbitrary code and exfiltrate data.
Users must upgrade to Cloud Run MCP version 1.11.0 or later.
Executive ViewAI
CVE-2026-85297 is a high-severity bucket squatting vulnerability in Google Cloud Run MCP versions before 1.11.0 that allows an unauthenticated attacker to intercept proprietary source code and tamper with deployments, potentially leading to arbitrary code execution and data exfiltration. While exploitation requires specific preconditions and no public exploit has been identified, successful exploitation could compromise the integrity of software deployments and expose sensitive intellectual property, making it a leadership concern for organizations using this tool.
Apply vendor patch.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
EUVD-2026-109906
GHSA-cg86-fr78-hv98
NVD
CVE.org
EPSS
Exploit-DB