Full Report
An amended complaint identifies a number of overseas individuals as key players “at the center of a global cybercrime network” that sold access to jailbroken generative AI tools. The post Microsoft IDs developers behind alleged generative AI hacking-for-hire scheme appeared first on CyberScoop.
Analysis Summary
# Threat Actor: Storm-2139
## Attribution & Identity
This threat network is tracked by Microsoft as **Storm-2139**. Key identified players include:
* **Arian Yadegarnia** (aka “Fiz”) from **Iran**
* **Ricky Yuen** (aka “cg-dot”) from **Hong Kong**
* **Phát Phùng Tấn** (aka “Asakuri”) from **Vietnam**
* **Alan Krysiak** (aka “Drago”) from the **United Kingdom**
Microsoft also identified actors in Illinois and Florida but declined to name them pending further investigation. The group operates as a "global cybercrime network."
## Activity Summary
The primary activity involves running a **hacking-as-a-service operation** that sold access to **jailbroken generative AI tools** hosted on Microsoft Azure OpenAI. This access was sold to overseas parties who then used the compromised accounts to generate “harmful content.” The initial court action prompted internal panic, with members speculating on identities and posting personal information about the handling Microsoft lawyer. Some accused parties contacted Microsoft to shift blame to others.
## Tactics, Techniques & Procedures
- Used **stolen Microsoft API keys** to gain unauthorized access to Azure OpenAI services.
- Exploited generative AI tools by **bypassing safety guidelines** (jailbreaking) to generate prohibited content.
- Shared personal information and photos of the opposing counsel (Microsoft lawyer) on chat forums.
- Post-litigation activity involved attempts to **discredit or shift blame** onto other members of the network.
- *MITRE ATT&CK IDs were not explicitly mentioned in the source text.*
## Targeting
- **Sectors:** Not explicitly detailed beyond the use of **Generative AI/Cloud Services** (Azure OpenAI). Victims are those requiring generative AI services susceptible to abuse.
- **Geography:** The network leaders are based in **Iran, Hong Kong, Vietnam, and the United Kingdom**. Buyers and users of the service are described as being **overseas**.
- **Victims:** The primary victim appears to be **Microsoft/OpenAI** due to misuse of their infrastructure, and indirectly **celebrities or public figures** whose likenesses were potentially used in generated imagery.
## Tools & Infrastructure
- **Malware Families Used:** Not specified.
- **Infrastructure (C2, domains, IPs):**
- Mention of a **Discord server** allegedly run by Krysiak used to sell Azure access starting at over $100.
- Links to **GitHub pages** hosting their software were shared.
## Implications
This case highlights a professionalized, international cybercrime operation focused on monetizing access to powerful generative AI capabilities by circumventing platform safety guardrails. The identification of key developers across multiple jurisdictions presents a complex trans-national enforcement challenge. The group displayed concerning reactive behaviors post-legal action, including doxxing and blame-shifting, indicating a resilient and coordinated operation.
## Mitigations
- **API Key Security:** Stricter management and monitoring of Microsoft API keys, especially those granting access to sensitive services like Azure OpenAI.
- **Content Moderation/Abuse Detection:** Enhanced monitoring for policy-violating content generation or sustained attempts to jailbreak models.
- **Legal and Law Enforcement Cooperation:** Continued preparation and submission of **criminal referrals** to US and foreign law enforcement bodies to disrupt international operations.
- **Insider/Associate Awareness:** Vigilance regarding threat actors attempting to coordinate disinformation or shift blame following legal action.