Loading story…
Palo Alto Networks Unit 42 reported an active supply-chain threat dubbed phantom squatting, in which large language models generate plausible but nonexistent brand-related domains that attackers later register and weaponize. In a study covering 913 global brands, 685,339 prompts across two LLM families produced 2.1 million URLs; researchers found 13,229 already classified as malicious and roughly 250,000 unregistered hallucinated domains that could be claimed by adversaries. The report says the technique creates a new path for phishing, malware delivery, and brand impersonation by exploiting domains that appear credible because they are syntactically consistent with legitimate organizations.
Unit 42 said defenders observed multiple real-world cases in which hallucinated domains were identified 18 to 51 days before attackers registered them, including a Montana Empire phishing operation and a postal-themed malicious Android APK campaign. The researchers warned that phantom-squatted domains can evade traditional reputation-based URL defenses because newly registered domains start with no threat history, while trusted AI assistants and autonomous agents may surface or use those domains directly, increasing software supply-chain and user-targeting risk.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
1 event from the most recent confirmed update back to the earliest known activity.
Unit 42 published research describing "phantom squatting," a threat in which attackers register and weaponize AI-hallucinated brand-related domains. The study analyzed 913 global brands across 685,339 prompts, generating 2.1 million URLs and identifying 13,229 already classified as malicious plus about 250,000 unregistered hallucinated domains.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
4 references tracked. Mallory keeps watching after this page renders.
scworld.com
Open sourcethehackernews.com
Open sourceunit42.paloaltonetworks.com
Open sourcedarkreading.com
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Threat actors are increasingly using **cybersquatting** (e.g., typosquatting, combosquatting, TLD squatting, and homograph domains) to impersonate legitimate brands and services for **credential phishing**, **malware delivery**, and fraud. Reporting cited WIPO handling a record ~6,200 domain disputes in 2025 (up 68% since 2020) and research indicating a sharp rise in malicious campaigns leveraging squatted domains, with the vast majority used to phish credentials or deliver malware; one example described impersonation of Decodo (formerly Smartproxy) via lookalike domains such as `smartproxy.org` and `smartproxy.cn`, which allegedly led to customer losses and reputational damage for the legitimate firm. Separately, incident response research described an **AI-assisted business impersonation** operation that registered **150+ domains spoofing law firms**, using multiple registrars and IP ranges, unique SSL/TLS certificates, and **Cloudflare** to conceal infrastructure. The operation appeared designed to target scam victims repeatedly (including reuse of a phone number tied to prior scam campaigns), and researchers warned that AI is lowering the barrier for less-skilled actors to run more convincing phishing/social engineering and fraud operations at higher scale.
Jun 29, 2026
Security researchers reported an emerging web attack technique that uses **generative AI** to turn a benign webpage into a malicious phishing or credential-stealing page *at runtime*. In a proof of concept attributed to Palo Alto Networks’ **Unit 42**, a “clean” page embeds instructions that trigger calls to public LLM APIs (e.g., Google Gemini, DeepSeek) to generate malicious JavaScript after the victim loads the site; the code is then executed in the browser, leaving little or no static payload to detect. Because the generated content is fetched from trusted AI service domains, the approach can also reduce the effectiveness of some network filtering and static analysis controls. Separately, an Anthropic evaluation highlighted that modern AI models are increasingly capable of conducting **multi-stage network attacks** using only standard, open-source tooling rather than specialized custom toolkits. The write-up notes that Claude Sonnet 4.5 could, in some simulated environments, identify a known public CVE and produce working exploit code quickly enough to exfiltrate sensitive data in an Equifax-like breach simulation, underscoring how AI can compress attacker timelines and increase the importance of fundamentals such as rapid patching and vulnerability management.
Jun 29, 2026
Threat actors are increasingly using **trusted platforms**—including cloud hosting and major ad networks—to deliver multi-stage phishing and scam campaigns that evade traditional URL and domain reputation controls. Recent activity includes a **three-step malvertising chain** delivered via **Facebook paid ads** that redirects victims through a decoy site (e.g., a fake Italian restaurant page) before landing on a **tech support scam (TSS) kit** hosted on **Microsoft Azure** infrastructure (including `web.core.windows.net`). Researchers reported rapid infrastructure churn, with **100+ domains rotated in seven days**, and targeting focused on **U.S. users** with activity concentrated on weekdays. Parallel enterprise-focused campaigns are hosting phishing infrastructure on **Microsoft Azure Blob Storage**, **Google Firebase**, and **AWS CloudFront**, using **redirect chains, CAPTCHA gates, and QR codes** to bypass automated analysis and email defenses. Analysis highlighted the use of **Adversary-in-the-Middle (AiTM)** phishing-as-a-service kits—**Tycoon2FA**, **Sneaky2FA**, and **EvilProxy**—to steal credentials and **session tokens** even when MFA is enabled. Separately, researchers documented a “clean email” approach to steal **Dropbox** credentials: benign-looking procurement-themed emails deliver **PDF attachments** that hide clickable elements (e.g., via *AcroForms* and `FlateDecode`), which then route victims to a second-stage file hosted on **Vercel Blob** and ultimately to a fake Dropbox login page that captures credentials and collects victim telemetry (IP address, location, and device details).
Jun 29, 2026