# MessiahGPT: Uncensored Criminal AI Model Marketed on BreachForums for Malware, Phishing, and Fraud Generation

> Trellix Advanced Research Center identified MessiahGPT, a criminal AI-as-a-service model advertised on BreachForums since July 2026 and offered live at messiahgpt[.]de plus a Telegram community, claiming zero ethical constraints (no RLHF, no Constitutional AI layer) to generate ransomware, phishing kits, stealers, crypters, rootkits, exploits/PoC code, fraud content, and more for as little as $8/month in crypto.

- **Published:** 2026-08-16T00:00:00Z
- **Last reviewed:** 2026-08-16T00:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-2036
- **Card image:** https://intel.threadlinqs.com/og/threat/threat-intel-messiahgpt-ai-malware-service-tl-2026-2036.webp?v=3af3bd67 (Threadlinqs threat intel: MessiahGPT Malware-Generating AI Service: Detection & IOCs (TL-2026-2036) — High severity, 6 IOCs, 9 detection rules.)
- **ID:** TL-2026-2036
- **Severity:** HIGH
- **Category:** MALWARE
- **Status:** ACTIVE
- **Detections:** 9 (1 shown free below) · **IOCs:** 6

## Description

MessiahGPT is a criminal AI service openly marketed on BreachForums, first advertised in an English-language dark-web listing in July 2026 and subsequently documented by the Trellix Advanced Research Center, whose findings were reported by Cyber Security News on 2026-08-14. The operator claims the underlying model was 'trained from scratch' rather than jailbroken from an existing commercial model, asserting no Reinforcement Learning from Human Feedback (RLHF), no Constitutional AI safety layer, and no internal concept of harm or illegality. The operator further claims a Mixture-of-Experts architecture (128 total experts, 16 active per token) trained on 'unrestricted manuals, dark web archives, leaked documentation, and raw internet scrapes with no post-filtering.' Trellix explicitly notes these architecture and training claims cannot be independently verified.

The service advertises on-demand generation of ransomware, phishing kits, stealers, crypters, and rootkits, as well as social engineering scripts, fraud and carding guides, data breach exploitation content, and — beyond the cyber domain — physical and chemical/explosive attack planning material. Separately, Accenture and Google Threat Intelligence Group researchers (Ryan Whelan; John Hultquist), presenting related AI-abuse findings at Black Hat USA 2026 and covered by Cybersecurity Dive on 2026-08-12, cited MessiahGPT as a tool observed generating exploits, payloads, proof-of-concept code, and refactored/rewritten malware for buyers who previously needed genuine development skill or a malware-as-a-service relationship.

Access is frictionless and low-cost: a free tier offers 50 queries with no registration, and paid tiers start around $8/month, payable only in cryptocurrency with no KYC verification. The listing includes a benchmark comparison table pitting MessiahGPT against ChatGPT-4o, DeepSeek-V3, and Mistral-Large, marketing itself as the only model that returns usable output across every category the mainstream models refuse.

Trellix separately tracks a related/competing service, DarkGPT, circulating on Russian-language Telegram channels and marketed as 'BlackHat AI uncensored power for darknet projects,' offering three free queries before paid tiers. Trellix frames both as part of a broader 2026 shift in which uncensored AI-as-a-service has matured from informal Telegram bots into dedicated platforms with versioned websites, demo channels, and tiered subscription pricing — a standing criminal product category rather than a novelty.

Direct verification on 2026-08-16 found messiahgpt[.]de still live and reachable, now fronted with the headline 'MessiahGPT — AI for Coding, Pentesting & OSINT | 25 Free Msgs/hr' — a dual-use, legitimizing rebrand of the same underlying service, consistent with the operator's strategy of framing overtly offensive capability as generic security-research tooling to reduce takedown/abuse-report friction. No confirmed intrusion, breach, or malware sample has yet been publicly attributed to MessiahGPT-generated output in the sources reviewed; the threat as documented is the existence and low-barrier availability of the generation service itself.

## MITRE ATT&CK

- T1593 Search Open Websites/Domains
- T1587 Develop Capabilities
- T1587.001 Malware
- T1587.004 Exploits
- T1588.007 Artificial Intelligence
- T1583.001 Domains
- T1566 Phishing
- T1204 User Execution
- T1027 Obfuscated Files or Information
- T1027.002 Software Packing
- T1014 Rootkit
- T1005 Data from Local System
- T1657 Financial Theft

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**SPL — Network Access to MessiahGPT Criminal AI-as-a-Service Infrastructure** (`TL-2026-2036-SPL-001`)

Detects DNS/proxy telemetry for corporate assets reaching messiahgpt[.]de, BreachForums, or MessiahGPT/DarkGPT Telegram channels — the criminal AI-as-a-service platform Trellix documented generating ransomware, phishing kits, stealers, crypters, and rootkits for $8/month with no KYC. Flags both attacker reconnaissance/capability-acquisition traffic and insider misuse of the service.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1593, T1588.007, T1583.001, T1566, T1204
- **Data sources:** proxy_logs, dns_logs, network_traffic
- **Severity:** high
- **Confidence:** high

### Splunk SPL

```spl
index=proxy (sourcetype=proxy_access OR sourcetype=stream:dns)
| eval req_domain=coalesce(dest_domain, query, replace(url,"^\w+://([^/]+).*","\1"))
| where match(req_domain, "(?i)messiahgpt\.de") OR match(req_domain, "(?i)breachforums\.(st|is|cx|to)") OR match(url, "(?i)t\.me\/(messiahgpt|darkgpt)")
| eval infra_category=case(match(req_domain,"(?i)messiahgpt"),"criminal_ai_saas_access", match(req_domain,"(?i)breachforums"),"dark_web_marketplace", match(url,"(?i)t\.me"),"telegram_criminal_ai_channel")
| stats count earliest(_time) as first_seen latest(_time) as last_seen values(src_ip) as src_ips values(user) as users by req_domain infra_category
| where count >= 1
| sort - count
```

### Microsoft KQL

```kql
DeviceNetworkEvents
| where Timestamp > ago(24h)
| where RemoteUrl has_any ("messiahgpt.de", "breachforums") or (RemoteUrl has "t.me" and (RemoteUrl has "messiahgpt" or RemoteUrl has "darkgpt"))
| extend InfraCategory = case(RemoteUrl has "messiahgpt", "criminal_ai_saas_access", RemoteUrl has "breachforums", "dark_web_marketplace", "telegram_criminal_ai_channel")
| summarize Count=count(), FirstSeen=min(Timestamp), LastSeen=max(Timestamp), Devices=make_set(DeviceName) by RemoteUrl, InfraCategory
| where Count >= 1
```

### Sigma

```yaml
title: Network Access to MessiahGPT Criminal AI-as-a-Service Infrastructure
id: 7a1f3b2e-9c4d-4e6f-8a2b-1d5e6f7a8b90
status: experimental
description: Detects DNS or proxy telemetry for connections to messiahgpt[.]de, BreachForums, or MessiahGPT/DarkGPT Telegram channels — criminal AI-as-a-service platforms advertising on-demand ransomware, phishing-kit, stealer, crypter, and rootkit generation for $8/month with no KYC.
references:
    - https://cybersecuritynews.com/messiahgpt-ai-tool/
    - https://www.trellix.com/news/press-releases/trellix-uncovers-diversification-of-ransomware-ecosystem-as-cybercriminal-use-of-ai-expands/
author: Threadlinqs Team
date: 2026-08-16
tags:
    - attack.reconnaissance
    - attack.t1593
    - attack.resource-development
    - attack.t1588.007
    - attack.t1583.001
    - attack.initial-access
    - attack.t1566
logsource:
    category: proxy
    product: generic
detection:
    selection_domain:
        c-uri|contains:
            - 'messiahgpt.de'
            - 'breachforums'
    selection_telegram:
        c-uri|contains:
            - 't.me/messiahgpt'
            - 't.me/darkgpt'
    condition: 1 of selection_*
falsepositives:
    - Authorized threat-intel researchers or SOC analysts actively monitoring the MessiahGPT/BreachForums ecosystem
level: high
```

**Known false positives:**

- Authorized threat-intel/SOC researchers actively monitoring MessiahGPT/BreachForums/DarkGPT for intelligence purposes
- Automated web-scraping or brand-monitoring tooling with broad crawl scope

**Tuning notes:** Expand the domain/URL indicator list as MessiahGPT/DarkGPT stand up mirror domains or rebrand — the operator was already observed on 2026-08-16 rebranding messiahgpt[.]de as 'AI for Coding, Pentesting & OSINT'; allowlist verified TI-research egress IPs to reduce analyst-noise false positives.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- KQL — Sentinel Detection of MessiahGPT/BreachForums Criminal AI Infrastructure Access (locked: unlock with TL Creds)
- KQL — Sentinel Detection of AI-Generated Crypter/Rootkit Defense-Evasion Staging (locked: unlock with TL Creds)
- KQL — Sentinel Detection of Ransomware Encryption and Stealer Credential/Wallet Access (locked: unlock with TL Creds)
- Sigma — Proxy/DNS Telemetry for MessiahGPT Criminal AI-as-a-Service Access (locked: unlock with TL Creds)
- Sigma — Unsigned Packed Dropper with Ransomware-Prep Commands (AI-Generated Crypter/Rootkit Staging) (locked: unlock with TL Creds)
- Sigma — Mass File Encryption and Stealer Credential/Wallet Access (AI-Generated Ransomware/Stealer Impact) (locked: unlock with TL Creds)
- SPL — Unsigned Packed Dropper Execution with Ransomware-Prep Commands (AI-Generated Crypter/Rootkit Staging) (locked: unlock with TL Creds)
- SPL — Mass File Encryption, Ransom-Note Drop, or Stealer Credential/Wallet Access (AI-Generated Ransomware/Stealer Impact) (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sources

- [MessiahGPT AI Tool Generates Malware and Phishing Content Without Any Restrictions](https://cybersecuritynews.com/messiahgpt-ai-tool/)
- [Researchers find AI-powered hacking tools for sale in underground forums](https://www.cybersecuritydive.com/news/ai-hacking-tools-sale-underground-forums/827807/)
- [Hackers abuse AI models to find new entry paths](https://www.cybersecuritydive.com/news/google-cloudaccenture/827689/)
- [Hackers abuse AI models to find new entry paths (CIO Dive)](https://www.ciodive.com/news/google-cloud-accenture-AI-hacks/827735/)
- [Hackers Using New BlackHat AI Tool MessiahGPT to Generate Ransomware and Phishing Kits](https://www.cryptika.com/hackers-using-new-blackhat-ai-tool-messiahgpt-to-generate-ransomware-and-phishing-kits/)
- [Trellix and BreachForums: Hackers Using New Blackhat AI Tool MessiahGPT to Generate Ransomware and Phishing Kits](https://blog.rankiteo.com/undtre1786703238-trellix-breachforums-cyber-attack-august-2026/)
- [Trellix Uncovers Diversification of Ransomware Ecosystem as Cybercriminal Use of AI Expands](https://www.trellix.com/news/press-releases/trellix-uncovers-diversification-of-ransomware-ecosystem-as-cybercriminal-use-of-ai-expands/)
- [MessiahGPT platform (live observation)](https://messiahgpt.de/)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-2036
