Severity: Critical (CVSS 10)

Affected Systems: Affected products listed by OpenCVE and the vendor advisory

Overview

Decepticon: Role-boundary forgery via ChatML special-token literals in web crawl output...

A critical vulnerability identified as CVE-2026-61732 has been disclosed.

Decepticon is an autonomous hacking agent for red teams. Versions prior to 1.1.17 wrap web crawl results — the output of agent reconnaissance against target services — into LLM messages without neutralizing ChatML special-token literals. Under the BYOK (Bring Your Own Key) deployment model, users configure their own LLM credentials to any OpenAI-compatible endpoint. Most open-source and self-deployed model providers (vLLM, SGLang, Ollama, LM Studio, text-generation-webui, etc.) do not filter special-token literals from user content in their default configurations. Those literals are parsed into structural role-boundary token IDs, meaning an attacker string planted in a target web page forges a new operator turn the model treats as authoritative, bypassing Decepticon's agent guardrails and resulting in arbitrary command execution inside the Kali Linux sandbox. Version 1.1.17 patches the issue.

Risk

CVSS and CISA data indicate the following:

  • Review the OpenCVE and vendor advisory for exploit conditions and impact

OpenCVE Analysis

CVSS v4.0 N/ACVSS v3.1 10 CriticalCVSS v3.0 N/ACVSS v2 N/AKEV noEPSS noSSVC no

  • OpenCVE title: Decepticon: Role-boundary forgery via ChatML special-token literals in web crawl output composed into LLM context
  • Severity score: Critical (CVSS 10)
  • Weaknesses: CWE-74
Attack VectorNetwork
Attack ComplexityLow
Privileges RequiredNone
User InteractionNone
ScopeChanged
Confidentiality ImpactHigh
Integrity ImpactHigh
Availability ImpactHigh
Attack VectorAttack ComplexityPrivileges RequiredUser InteractionScopeConfidentiality ImpactIntegrity ImpactAvailability Impact

Required Action

Review the OpenCVE detail page and linked vendor advisory, then apply the vendor-provided update or mitigation for the affected product.

Prioritize systems where the affected product is internet-facing, handles authentication, or runs with elevated privileges.

Verify Updates

Confirm whether your environment uses the affected product(s): Affected products listed by OpenCVE and the vendor advisory.

After remediation, verify the installed version against the fixed or unaffected versions listed by the vendor.

Temporary Mitigation (if patch is not available)

Use the mitigation published by the vendor. If no vendor mitigation is available, reduce exposure to the affected product, restrict access to trusted users or networks, and increase monitoring until an update can be applied.

Recommendation

  • Use OpenCVE, vendor, and source references as the source of truth for affected versions and remediation
  • Patch or mitigate affected products after confirming exposure in your environment
  • Monitor affected systems for unusual activity until remediation is complete

Support

If you require assistance, please contact our support team.

Immediate action is strongly recommended to protect your infrastructure.

Source Details

Customer Responsibility and Backups

Before applying updates, mitigations, or configuration changes, customers should take and verify current backups or snapshots of affected systems.

Customers are responsible for managing their servers, validating their own backups, testing changes, and ensuring they can restore services if an update or mitigation causes an issue.



Thursday, September 24, 2026

« Back