AI-assisted phishing kit targets Microsoft Outlook users

Researchers say a modular phishing operation is harvesting credentials through fake Microsoft login pages and rotating exfiltration channels.

What happened

Security researchers identified a phishing campaign targeting Microsoft Outlook users, primarily Spanish-speaking victims, using a credential harvesting kit known as Mycelial Mage. According to reporting by Cyber Press, the operation mimics Microsoft login pages and captures usernames, passwords, IP addresses, and geolocation data. Researchers traced early activity to March 2025, with broader analysis beginning later in the year as additional domains and variants were uncovered.

Going deeper

The phishing kit relies on modular JavaScript components that allow operators to change command and control endpoints without altering the core phishing logic. Early versions transmitted stolen data through Telegram bots using lightly obfuscated tokens embedded in script files. Later versions introduced more defensive code, including debugger detection and runtime interference, which complicated analysis. More recent variants shifted exfiltration to Discord webhooks, which operate as one-way channels and prevent defenders from retrieving previously stolen data. Across all versions, the workflow remained consistent, with credentials enriched using external IP and location services before being transmitted in a standardized format. Researchers noted that the reuse of code structure, markers, and infrastructure points to a service style operation rather than isolated attacks.

What was said

Analysts said the campaign shows deliberate engineering choices designed to support scale and reuse. They observed consistent formatting, structured functions, and clear documentation in later scripts, suggesting automated assistance in development rather than ad hoc scripting. Researchers also indicated that rotating between messaging platforms reduced exposure to monitoring, as some services allow defenders to replay stolen data if access tokens are recovered. The use of disposable domains and interchangeable exfiltration channels was described as a way to sustain operations even when individual components are taken down.

The big picture

Microsoft says campaigns like this indicate a shift toward AI-assisted phishing operations that are built for speed and scale. In its 2025 Digital Defense Report, the company warns that adversaries are now using generative AI to “scale social engineering” and automate parts of the attack chain that once required hands-on effort. Microsoft notes that “autonomous malware and AI-powered agents are now capable of adapting their tactics on the fly,” making modular phishing kits more resilient when domains, scripts, or exfiltration paths are disrupted.

The report also points to AI’s growing part in making phishing harder to spot. Microsoft says the “integration of generative AI into adversarial operations has significantly elevated the persuasiveness and scale of social engineering campaigns,” and expects attackers to rely even more heavily on these techniques as organizations strengthen traditional defenses. That shift aligns with phishing kits like Mycelial Mage, where reusable components and rotating infrastructure allow operators to sustain credential theft with minimal friction and rapid recovery from takedowns.

FAQs

What makes this phishing kit different from older campaigns?

It uses interchangeable components and rotating exfiltration channels, which allows operators to update infrastructure without rebuilding the phishing pages.

Why are Outlook users frequently targeted?

Email accounts often provide access to other services, internal communications, and password reset workflows, making them valuable entry points.

How does rotating between Telegram and Discord help attackers?

Different platforms offer varying levels of visibility to defenders. One way webhook channels limit the ability to inspect stolen data after exfiltration.

Does encryption prevent these attacks?

Encryption protects data in transit but does not stop users from submitting credentials to fake login pages that appear legitimate.

How can organizations reduce risk from phishing kits?

They can enforce phishing resistant authentication, limit browser based credential reuse, train users to verify login URLs, and monitor for lookalike domains.