Skip to navigation

Threat reportSupply ChainTL-2026-1050

Phantom Squatting: Adversaries Preemptively Register AI-Hallucinated Domains to Hijack Software Supply Chain Trust (Unit 42 "Montana Empire" Case)

highACTIVE

Unit 42 research shows LLMs consistently hallucinate plausible-but-nonexistent brand domains, and adversaries are preemptively registering these hallucinated domains to intercept traffic from humans and autonomous agents that trust AI-generated URLs.

Key facts

IdentifiersTL-2026-1050
Also known asPhantom Squatting, Montana Empire
SeverityHIGH
Affected productsMultiple LLM-generated brand/domain output (production-optimized mini-class enterprise model, low-latency lite-class frontier model); Various 913 global brands across technology, finance, healthcare, e-commerce, government, gambling, and logistics sectors
MalwareMontana Empire, PhantomRaven
ToolsAI coding assistant (unspecified)
MITRE ATT&CK techniquesT1005 Data from Local System, T1102 Web Service, T1102.002 Bidirectional Communication, T1111 Multi-Factor Authentication Interception, T1189 Drive-by Compromise, T1204.001 Malicious Link, T1204.002 Malicious File, T1539 Steal Web Session Cookie, T1566 Phishing, T1566.002 Spearphishing Link, T1567 Exfiltration Over Web Service, T1583.001 Domains and 9 more in the attack flow
MotivationFINANCIAL
Target sectorstechnology, finance, health, ecommerce, government administration, gambling, logistics, postal-services, retail-banking
Target regionsGlobal, turkey, united arab emirates, bangladesh, Europe
Detection rules9 (1 shown free)
Sources10 cited sources

Detection coverage

As of 2026-07-01, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-1050 across Splunk SPL, Microsoft KQL and Sigma, covering 15 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.

Free detection: User Navigation to Newly-Registered Zero-Reputation Domain Resembling Monitored Brand (Phantom-Squatting Landing)

Detects DNS/proxy telemetry where an endpoint resolves or browses to a domain that lexically resembles a watchlisted brand string (path/subdomain/root-level lookalike patterns such as 'secure-<brand>', '<brand>-portal', '<brand>-verify') and carries a WHOIS registration age under 60 days with no established reputation score — the exploitation entry point for the Montana Empire-style phantom-squatting phishing kit described in research.json.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1594, T1566.002, T1583.001
  • Data sources: dns_logs, proxy_logs, threat_intelligence_feed
  • Severity: high
  • Confidence: medium

Splunk SPL

index=proxy sourcetype="stream:dns" OR sourcetype="pan:threat" OR sourcetype="symantec:proxysg"
| eval domain_age_days=coalesce(domain_age_days, whois_age_days)
| where (match(query, "(?i)secure-") OR match(query, "(?i)-portal") OR match(query, "(?i)-verify") OR match(query, "(?i)-login"))
| where domain_age_days<60
| where reputation_score=0 OR isnull(reputation_score) OR reputation="unknown"
| stats count min(_time) as first_seen max(_time) as last_seen by src_ip, dest_ip, query, domain_age_days, reputation_score
| where count>0
| table first_seen, last_seen, src_ip, dest_ip, query, domain_age_days, reputation_score, count

Microsoft KQL

let brandTerms = dynamic(["secure-", "-portal", "-verify", "-login"]);
DnsEvents
| where TimeGenerated > ago(24h)
| where Name has_any (brandTerms)
| join kind=leftouter (ThreatIntelligenceIndicator | project NetworkDomain, ConfidenceScore) on $left.Name == $right.NetworkDomain
| extend DomainAgeDays = todynamic(AdditionalFields).domain_age_days
| where isempty(ConfidenceScore) or ConfidenceScore == 0
| where todouble(DomainAgeDays) < 60 or isempty(DomainAgeDays)
| project TimeGenerated, Computer, ClientIP, Name, DomainAgeDays, ConfidenceScore
| summarize FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated), Count=count() by Computer, ClientIP, Name

Sigma

title: Endpoint Contacts Newly Registered Zero-Reputation Brand-Lookalike Domain
id: 7c1e2a4a-8b3d-4f1e-9a6c-1050a0011001
status: experimental
description: Detects DNS queries or proxy connections to domains that lexically resemble a monitored brand (path/subdomain/root-level lookalike fabrications) with a WHOIS age under 60 days and no threat-intel reputation history, consistent with the phantom-squatting / Montana Empire attack pattern.
references:
    - https://unit42.paloaltonetworks.com/phantom-squatting-hallucinated-web-domains/
    - https://thehackernews.com/2026/07/phantom-squatting-uses-ai-hallucinated.html
author: Threadlinqs Team
date: 2026-07-01
tags:
    - attack.reconnaissance
    - attack.t1594
    - attack.resource_development
    - attack.t1583.001
    - attack.initial_access
    - attack.t1566.002
logsource:
    category: dns_query
detection:
    selection_lookalike:
        query|contains:
            - 'secure-'
            - '-portal'
            - '-verify'
            - '-login'
    selection_zero_rep:
        answer_reputation: 'unknown'
    condition: selection_lookalike and selection_zero_rep
falsepositives:
    - Legitimate marketing/staging subdomains recently registered by the actual brand owner
    - Newly onboarded SaaS vendor domains with similar naming conventions
level: high

Known false positives:

  • Legitimate marketing/staging subdomains recently registered by the actual brand owner
  • Newly onboarded third-party vendor domains with similar naming conventions

Tuning notes: Seed the lookalike term list from an organization-specific hallucination-surface mapping exercise (per research.json remediation) rather than generic strings; tighten domain_age_days threshold to the org's own observed AEW distribution (18 minutes–51+ days) and require reputation_score absence AND lookalike match to reduce noise from legitimate new vendor domains.

8 more detection rules (Splunk SPL, Microsoft KQL, Sigma) · unlock with TL Creds. Sign up free to unlock the detection rules or query them with the Threadlinqs MCP server (Purple)
  • Autonomous Agent / Browser Auto-Navigation to LLM-Suggested URL Without Allowlist Validation · unlock with TL Creds
  • Registry/Config Artifact Persists Reference to Brand-Lookalike Phantom Domain String · unlock with TL Creds
  • Malicious Sideloaded APK Install Following Cloned-Storefront Page View (Mobile Phantom-Squatting Distribution) · unlock with TL Creds
  • Coordinated Bulk Domain-Registration Cluster for Brand-Adjacent Strings (Automated Phantom-Domain Acquisition) · unlock with TL Creds
  • AI Coding Assistant Session Artifacts Consistent with Phishing-Kit Development (Insider/Compromised-Dev Host) · unlock with TL Creds
  • Outbound Telegram Bot API Call Immediately Following Web-Form Submission on Endpoint or Web Server (C2/Exfil) · unlock with TL Creds
  • Credential/OTP Form Field Values Relayed via PowerShell Invoke-WebRequest Shortly After Page Render · unlock with TL Creds
  • Ad-Hoc PHP Built-in Server Spawned Outside Dev Paths Serving Brand-Impersonation Content (C2/Exfil Relay) · unlock with TL Creds

Indicators of compromise

As of 2026-07-01, Threadlinqs Intelligence tracks 15 indicators of compromise for TL-2026-1050.

Sample indicators (2 of 15)

  • File hash: eb07edaa2786cfddfa4c15526168f2200d85300aee0a8f253b32d2462a7b0bcd
  • File hash: 2202a30daad9928ef47cca5f4ab04ce083692a94428e386fa01c2dd44557e34b

Defanged, TLP:CLEAR. Full set of 15 indicators: unlock with TL Creds.

15 indicators of compromise · unlock with TL Creds. Compare plans

Technical analysis

Analysis of 913 global brands across 685,339 adversarial prompts (2.1M unique generated URLs) found 13,229 confirmed malicious URLs and ~250,000 unregistered phantom domains still available for takeover; the documented "Montana Empire" case shows a hallucinated postal-ecommerce domain flagged 23 days before an attacker registered it and stood up an AI-assisted, PHP-based credential-phishing kit with Telegram C2.

Unit 42 (Palo Alto Networks) coined the term "phantom squatting" to describe a software supply-chain attack vector in which large language models (LLMs) hallucinate plausible but non-existent domains when asked about legitimate brands, and threat actors preemptively or reactively register those exact hallucinated domains to weaponize them for phishing, malware distribution, and command-and-control before the brand or defenders can claim them.

Researchers tested two production LLMs — an unnamed 'production-optimized mini-class enterprise model' (dated April 2025) and a 'low-latency lite-class frontier model' (dated June 2025) — by issuing 685,339 adversarial brand-related prompts covering 913 global brands across technology, finance, healthcare, e-commerce, government, gambling, and logistics sectors. Prompts were run at three temperature settings (Precise T=0.1, Balanced T=0.7, Creative T=1.5) to test hallucination sensitivity to model randomness. The corpus of 2.1 million unique generated URLs was then checked against domain registration data and threat intelligence feeds.

Key findings: the first model hallucinated non-existent domains (NXD) at a 44.6% rate and the second at 27.5%; the Creative temperature setting produced the highest NXD rate (43.10%) confirming that higher-randomness generation increases hallucination risk. Of the full corpus, 13,229 URLs (0.61%) were confirmed malicious and 41,313 (1.90%) were high-risk; the confirmed-malicious set broke down as 67.2% malware, 16.2% phishing, 13.7% grayware, and 3.0% command-and-control. Roughly 250,000 hallucinated domains had no registered owner at time of analysis — a live, discoverable attack surface for adversaries to claim.

URL hallucinations were most commonly path-level fabrications on otherwise legitimate domains (49.7%), followed by subdomain-level fabrications (39.5%) and pure fabricated root domains (10.8%).

The flagship case study, 'Montana Empire,' involved a postal/e-commerce brand: Unit 42's multi-agent discovery pipeline flagged a hallucinated brand domain as a high-risk target on March 8, 2026. On March 31, 2026 — an adversarial exploitation window (AEW) of 23 days — an attacker registered the exact domain and deployed a phishing kit named 'Montana Empire.' The kit's distribution ZIP archive (SHA-256 eb07edaa2786cfddfa4c15526168f2200d85300aee0a8f253b32d2462a7b0bcd, 7,958,528 bytes) contained a PHP backend that served a real-time scraped clone of the legitimate storefront, harvested credentials, credit-card numbers, IBAN/bank-transfer details, and national identity documents, and relayed one-time passwords (OTPs) in real time. Stolen data and operator commands were exfiltrated and controlled via a Telegram bot serving as the kit's command-and-control channel. The kit's admin panel displayed the banner 'Kimseye Güvenme' ('Trust No One' in Turkish). Forensic artifacts (project files and session logs) indicated the operator used an AI coding assistant to build the phishing kit itself, layering AI-assisted attacker tooling on top of an AI-hallucination-driven target selection process.

A second documented case involved a national postal service brand clone distributing a malicious Android APK (SHA-256 2202a30daad9928ef47cca5f4ab04ce083692a94428e386fa01c2dd44557e34b) via a pixel-perfect cloned storefront that displayed a fabricated 4.8-star rating and false '2M+ users' claim; this case had a 51-day adversarial exploitation window. Additional detected phantom-squatting incidents targeted a Bangladesh-focused sports-betting brand (45-day AEW, credential harvesting), a second sports-betting brand with a coordinated dual domain registration 18 minutes apart (40-day AEW), a European retail bank (35-day AEW, re-registration event), and a UAE commercial bank domain that had been registered and abused roughly 11 months prior to detection, validating the historical persistence of the technique.

Unit 42 frames the vulnerability as structural and "inherently unpatchable": hallucinated domains are functionally indistinguishable from legitimate new domains at registration time, carry no prior threat-intelligence history, have no established reputation score, and are absent from all blocklists — allowing attackers to bypass reputation- and blocklist-based defenses entirely during the exploitation window between hallucination-discovery and domain weaponization. The vector is explicitly called out as a software supply-chain risk because both human users trusting AI chatbot/search answers and autonomous AI agents that programmatically follow LLM-generated URLs (e.g., in agentic browsing, coding assistants, or automated procurement workflows) are exposed to identical risk, without any code, package, or dependency being compromised. The report also connects phantom squatting to the related and independently reported 'slopsquatting' vector, in which code-generating LLMs hallucinate non-existent software package names (e.g., npm/PyPI) that attackers then register with malicious payloads; a related campaign, PhantomRaven, embedded malware in 126 npm packages that together achieved 86,000+ installs, and a documented case involved an attacker registering a hallucinated shortened package name ('unused-imports') derived from the legitimate 'eslint-plugin-unused-imports'.

Palo Alto Networks lists its own product mitigations (Advanced WildFire, Advanced URL Filtering, Advanced DNS Security, Prisma AIRS, Koi Agentic Endpoint Security, Unit 42 AI Security Assessment) and recommends organizational controls: proactively mapping an organization's own LLM hallucination surface before deploying AI assistants/agents, establishing phantom-domain watchlists tied to brand and product names, building continuous discovery pipelines that regenerate and re-check hallucinated URLs over time (since AEWs of 20-50+ days were observed), and requiring independent verification of any URL surfaced by an AI system before a human or autonomous agent acts on it (credential entry, file download, payment).

MITRE ATT&CK techniques by tactic:

Collection

T1005 Data from Local System

Command and Control

T1102 Web Service; T1102.002 Bidirectional Communication

Credential Access

T1111 Multi-Factor Authentication Interception; T1539 Steal Web Session Cookie

Initial Access

T1189 Drive-by Compromise; T1566 Phishing; T1566.002 Spearphishing Link

Execution

T1204.001 Malicious Link; T1204.002 Malicious File

Exfiltration

T1567 Exfiltration Over Web Service

Resource Development

T1583.001 Domains; T1583.006 Web Services; T1584.001 Domains; T1585.001 Social Media Accounts; T1587.001 Malware; T1608.001 Upload Malware

Reconnaissance

T1589 Gather Victim Identity Information; T1589.001 Credentials; T1594 Search Victim-Owned Websites

Impact

T1657 Financial Theft

Remediation

Immediate actions

  • Register defensive/watchlist domains for high-value hallucination-prone brand variants identified via LLM prompting
  • Block or warn on newly-registered domains (WHOIS age < 60 days) that resemble owned brand names before allowing outbound navigation
  • Deploy DNS security / URL filtering that flags zero-reputation domains resembling monitored brand strings
  • Alert users and internal AI agents against entering credentials or OTPs on any AI-suggested link without independent verification

Workarounds

  • Treat all LLM-generated URLs and package/domain names as unverified draft output requiring independent confirmation against an authoritative source before use
  • Prevent automatic OTP relay or credential auto-fill on domains not present on an internally maintained allowlist

Longer-term hardening

  • Map the organization's own brand hallucination surface by systematically prompting production LLMs and logging generated domains/URLs
  • Build a continuous discovery and re-check pipeline for hallucinated domains, since adversarial exploitation windows of 18 minutes to 51+ days were observed
  • Restrict autonomous AI agents (browsing, coding assistants, procurement bots) from auto-navigating to or auto-trusting LLM-generated URLs without allowlist validation
  • Extend brand-protection / anti-phishing monitoring programs to include LLM-hallucination-derived domain permutations, not just traditional typosquatting

Affected and fixed versions

  • Multiple — LLM-generated brand/domain output (production-optimized mini-class enterprise model, low-latency lite-class frontier model)
    Vulnerable versions: April 2025 model snapshot; June 2025 model snapshot
  • Various — 913 global brands across technology, finance, healthcare, e-commerce, government, gambling, and logistics sectors
    Vulnerable versions: N/A - structural LLM hallucination issue, not a software version defect

Timeline

  • LLM1 (production-optimized mini-class enterprise model) snapshot dated used in Unit 42 hallucination study
  • LLM2 (low-latency lite-class frontier model) snapshot dated used in Unit 42 hallucination study
  • Unit 42 multi-agent discovery pipeline generates and flags the hallucinated postal/e-commerce brand domain later used in the Montana Empire case as high-risk
  • Attacker registers the flagged hallucinated domain and deploys the Montana Empire PHP phishing kit with Telegram C2, 23 days after Unit 42's initial detection (adversarial exploitation window)
  • Infosecurity Magazine reports on related 'slopsquatting' AI package-hallucination supply chain threat
  • Cloud Security Alliance publishes research note on slopsquatting AI supply chain attacks
  • Dark Reading, GBHackers, CyberPress, Cybernews, and IT Security News publish coverage summarizing the Unit 42 findings
  • Unit 42 (Palo Alto Networks) publishes 'Phantom Squatting: AI-Hallucinated Domains as a Software Supply Chain Vector' research report, coining the term and detailing the 685,339-prompt / 913-brand / 2.1M-URL study
  • The Hacker News publishes detailed technical coverage including additional case breakdowns and recommended defenses

Sources cited

Previous and next in supply chain

Further reading

Drafted by the Threadlinqs research pipeline from 10 sources; published ; last updated —.

Threadlinqs Intelligence — Real-Time Threat Detection Platform

// threat_feed
$ sort --newest
Showing all threats

Live intelligence console

Threat level

Latest Threats