Threat reportThreat IntelligenceTL-2026-3813
Criminal Adoption of Unrestricted AI Services: MessiahGPT, Kriminal.ai, WormGPT-linked Actors and AI-Driven KYC Bypass (Flashpoint, Aug 2026 data)
Criminal Adoption of Unrestricted AI Services (TL-2026-3813), also tracked as MessiahGPT, is a medium-severity tracked intrusion set, first published 2026-10-17. It has no confirmed attribution, affects Multiple Identity verification (KYC) and onboarding workflows, maps to 11 MITRE ATT&CK techniques (T1078, T1195, T1566), and is covered by 9 detection rules and 15 indicators of compromise.
- Severity
- MEDIUMAssessed severity
- CVEs
- 0None referenced
- Techniques
- 11MITRE ATT&CK
- Actors
- 0Not attributed
- Detection rules
- 9SPL · KQL · Sigma
- IOCs
- 15Indicators of compromise
Key facts for TL-2026-3813
- Threat ID
- TL-2026-3813
- Also known as
- MessiahGPT, Kriminal.ai, Criminal AI services
- Severity
- MEDIUM
- Status
- ACTIVE
- Category
- THREAT_INTEL
- First published
- Last reviewed
- Attribution confidence
- LOW
- Motivation
- FINANCIAL
- Target sectors
- finance, technology, ecommerce, cryptocurrency, general
- Target regions
- Global
- Detection rules
- 9
- Indicators of compromise
- 15
Malware and tooling in Criminal Adoption of Unrestricted AI Services
Malware and tooling: Crypter, GROK, Kriminal, MessiahGPT, WormGPT, telegram, LiteLLM, OpenRouter
How Criminal Adoption of Unrestricted AI Services works
Flashpoint reports 7,461,908 illicit-community posts discussing AI in August 2026 (+34% vs July), including 751,644 posts on AI-enabled KYC bypass (+91%). Unrestricted LLM services such as MessiahGPT and Kriminal.ai are marketed for malware development, phishing kits and social engineering scripts, with distribution dominated by Telegram.
Flashpoint's monthly AI threat report (published 2026-10-06) measured 7,461,908 posts discussing AI across illicit communities in August 2026, up nearly 34% from 5,575,564 in July. Posts about AI-enabled identity-verification (KYC) bypass rose about 91%, from 394,567 to 751,644. Telegram dominated distribution with 4,899,407 posts advertising AI services, followed by GitHub Gist (20,382), Reddit (14,084) and Pastebin (9,487). Flashpoint lists identity-verification bypass, deepfakes, phishing, malware development, fraud and vulnerability analysis as the primary criminal uses, and notes that actors chain these capabilities into coordinated workflows (for example synthetic identities combined with compromised credentials).
MessiahGPT is an unrestricted LLM service that Flashpoint attributes to the operator 'dabial leaks' (associated with WormGPT). It was first observed in January 2026, offers a free tier plus plans of $9.99-$39.99/month, and is marketed for malware, phishing kits, social engineering scripts and analysis of breached data. Flashpoint cites an August 24 Pastebin post in which a user applied it to port-scan analysis. Separately, Trellix Advanced Research Center documented a BreachForums advertisement for MessiahGPT (reported by Hackread on 2026-08-18) with the site messiahgpt.de, 50 free queries without registration, paid plans from about $8/month, cryptocurrency-only payment with no identity verification, and claimed capabilities including ransomware, infostealers, crypters, rootkits, phishing kits and fraud/carding guidance. Trellix could not independently verify the claimed model architecture, training data or performance; these are operator claims.
Kriminal.ai is priced from $12.99/month (top GHOST tier $99) and advertises built-in jailbreak prompts that circumvent safety restrictions. ThreatDown's analysis (reported by CSO Online on 2026-08-20) found it has no proprietary foundation model: it routes requests to xAI Grok (primary engine), Anthropic Claude, Mistral Large and Llama 3.3 via OpenRouter, with Tavily for live web search. It is hosted on the clearnet behind Cloudflare/Google Cloud with NowPayments crypto checkout, and sells exploit development, OSINT dossiers, on-chain tracing, social engineering and code generation, plus named GHOST-tier personas (PHANTUM, ARCHITECT, ORACLE, WRAITH).
Flashpoint also references a LAPSUS$ claim of 4 TB allegedly stolen from AI recruiting platform Mercor (proprietary AI models, candidate information, client records) and an unnamed major U.S. enterprise cloud/AI access listing. Other reporting dates the Mercor incident to early April 2026 and links it to a supply-chain compromise of the open-source LiteLLM library; Mercor confirmed a security incident. Note the date discrepancy with the Flashpoint framing. This is a trend report: no CVEs, CVSS scores or network IOCs beyond the service domains are stated, and BeaconBeagle shows no C2 config match for messiahgpt.de. Defenders should treat it as context for expected increases in AI-assisted phishing, lower-skill malware development, KYC/account-opening fraud and faster exploitation triage.
MITRE ATT&CK techniques used in TL-2026-3813
Initial Access
T1078 Valid Accounts; T1195 Supply Chain Compromise; T1566 Phishing
Resource Development
T1585 Establish Accounts; T1587.001 Develop Capabilities: Malware; T1588.007 Obtain Capabilities: Artificial Intelligence
Reconnaissance
T1589.001 Gather Victim Identity Information: Credentials; T1595 Active Scanning; T1598 Phishing for Information
Impact
stealth
Affected products and versions in Criminal Adoption of Unrestricted AI Services
- Multiple — Identity verification (KYC) and onboarding workflows
Vulnerable versions: Workflows without liveness or document-forensics controls - Multiple — Email and messaging users targeted by AI-generated phishing and social engineering
Vulnerable versions: All
Remediation for Criminal Adoption of Unrestricted AI Services
Immediate actions
- Monitor illicit channels (Telegram, paste sites, forums) for brand, executive and customer-data mentions tied to AI-assisted fraud
- Block newly registered look-alike domains and review phishing filtering for fluent, localized, AI-generated lures
- Review KYC/onboarding fraud rules for deepfake, synthetic-identity and reused-credential patterns
Workarounds
- Rate-limit and alert on unusual external port-scan and reconnaissance activity against exposed services
Longer-term hardening
- Add liveness detection and multi-signal identity verification (device, behavior, document forensics) to onboarding flows
- Enforce phishing-resistant MFA and credential-exposure monitoring to counter synthetic identities combined with compromised credentials
- Track emerging criminal AI tools and services as part of regular threat-landscape review
- Inventory and pin AI/LLM dependencies (for example LiteLLM) and monitor them for supply-chain compromise
Timeline of Criminal Adoption of Unrestricted AI Services
- MessiahGPT first observed (January 2026) per Flashpoint, operated by 'dabial leaks' (associated with WormGPT); exact day not stated
- Mercor confirms a security incident tied to a LiteLLM supply-chain attack; LAPSUS$ claims ~4 TB of stolen data (Flashpoint's report lists the claim among August activity)
- Trellix-documented BreachForums advertisement for MessiahGPT (messiahgpt.de) reported: 50 free queries, paid plans from ~$8/month, crypto-only payment
- ThreatDown finds Kriminal AI is largely a jailbroken wrapper over Grok, Claude and OpenRouter-hosted models, priced from $12.99/month
- Pastebin post shows a user applying MessiahGPT to port-scan analysis (cited by Flashpoint)
- End of August 2026 measurement window: 7,461,908 illicit AI-related posts (+34% vs July) and 751,644 KYC-bypass posts (+91%)
- Flashpoint publishes its monthly AI Threat Report on AI use across illicit communities
Sources cited for Criminal Adoption of Unrestricted AI Services
- AI Threat Report: How Artificial Intelligence Is Used Across Illicit Communities
- MessiahGPT Criminal AI Service Advertised on BreachForums
- 'MessiahGPT' AI Service Promises Ransomware, Phishing Kits, and Malware
- Kriminal breaks out of Grok, Claude guardrails at $12.99
- Criminal AI tool Kriminal is mostly just Grok with a jailbreak, ThreatDown finds
- No-Filter 'Kriminal' AI Platform Raises Cybercrime Concerns
- Mercor AI Confirms Data Breach Following Lapsus$ Claims of 4TB Data Theft
- Mercor, a $10 billion AI startup, confirms major data breach (Fortune)
Detection coverage for TL-2026-3813
As of 2026-10-17, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-3813 across Splunk SPL, Microsoft KQL and Sigma, covering 15 indicator(s) of compromise. The whole corpus is readable without an account; a free account unlocks full detection query text in Splunk SPL, Microsoft KQL and Sigma; paid tiers add raw indicator values, correlation and the MCP server. Threadlinqs MCP server · View plans.