CRAFTY FALCON Weaponizes AI Against Financial Firms

Amber | Attack
CRAFTY FALCON Weaponizes AI Against Financial Firms | Hive Pro Threat Advisory
Date of publication
November 10, 2026
Admiralty code
A1
TA number
TA2026329
First seen
October 9, 2026
Targeted regions
Asia
Targeted platform
Cloud (AWS), Web Applications
Targeted products
AWS Systems Manager, AWS Lambda, PHP applications, Node.js applications, Payment and Disbursement APIs, Merchant Settlement APIs, Cryptocurrency Wallets
Targeted industries
Banking, Financial Services, Cryptocurrency, and Insurance
Threat actor
CRAFTY FALCON
Malware
ARTEX, PentAGI, Cairn, CyberStrikeAI

Summary

CRAFTY FALCON is a financially motivated threat actor that uses artificial intelligence to attack financial organizations in India at machine speed. It chains publicly available AI-driven offensive tools (ARTEX, PentAGI, Cairn and CyberStrikeAI) to run attack chains with little human effort.

Initial access comes through exposed testing and staging environments or newly disclosed PHP and Node.js flaws, followed by web shells and abuse of AWS Systems Manager and Lambda. The actor cashes out with Rs. 1 test payments followed by much larger unauthorized transfers, merchant settlement abuse and cryptocurrency wallet theft.

NameOriginMotiveTarget countriesTarget industry
CRAFTY FALCONUnattributedFinancial GainAsiaBanking, Financial Services, Cryptocurrency, and Insurance

Actor Details

#1

CRAFTY FALCON is a financially motivated threat actor that uses artificial intelligence to attack financial organizations in India at machine speed. Its targets include non-banking financial companies, digital lenders, fintechs, payment aggregators, and cryptocurrency exchanges. No country of origin or aliases have been attributed to the actor.

#2

The actor's edge comes from publicly available AI-driven offensive tools. ARTEX is a family of tools built on large language models, PentAGI is an autonomous multi-agent penetration testing system, Cairn is an attack search engine, and CyberStrikeAI is an offensive framework. Together, they let the actor run entire attack chains with little human effort, write and refine custom scripts on the fly, and adapt payloads to how each target responds.

#3

Initial access usually comes through internet-exposed testing and staging environments, or through newly disclosed flaws in PHP and Node.js applications that the actor exploits almost as soon as they go public. Once inside, it plants web shells and abuses AWS Systems Manager and Lambda to run commands, blending in with normal admin activity.

#4

From there, the actor moves into live production systems using shared credentials and tokens. It also tampers with client apps to send unauthorized API calls and routes traffic through trusted third-party services to slip past IP allow-lists.

#5

To cash out, the actor first sends test payments as small as Rs. 1 to confirm a payment route works, then follows with much larger unauthorized transfers. It also abuses merchant settlement systems and steals cryptocurrency wallet keys to drain funds.

#6

The actor strikes during weekends and holidays, rotates its infrastructure within hours to days, and has returned to organizations that did not fully clean up after earlier attacks. Successful fraudulent transfers have already been confirmed, although total losses have not been disclosed.


Recommendations

  1. 01

    Keep Watch Around the Clock, Especially on Weekends and Holidays

    This actor strikes when teams are off duty. Make sure security monitoring runs 24x7, add extra coverage during holidays and weekends, and treat unusual activity during these periods with higher urgency.

  2. 02

    Lock Down Cloud Admin Tools and Watch Them Closely

    Only approved accounts should be able to run remote commands or change serverless functions, such as AWS Systems Manager and Lambda. Give every account the minimum access it needs, block internet access from these functions by default, and alert immediately when new users, roles, or access keys appear.

  3. 03

    Keep Test Systems Completely Separate from Live Systems

    Testing, staging, and UAT environments must never share passwords, tokens, or network connections with production. Remove unnecessary internet exposure from test systems and alert on any attempt to reach production from them.

  4. 04

    Patch Fast and Protect Web Applications

    Apply security updates for internet-facing applications as quickly as possible, especially PHP and Node.js systems. Where patching takes time, use a web application firewall as a temporary shield, and set web folders to read-only so attackers cannot easily drop malicious files.

  5. 05

    Verify Every Payment Request and Spot "Test" Transactions

    Never trust the app alone. Check every transaction on the server side, and replace IP-based trust with stronger checks such as signed requests or mutual authentication. Flag tiny Rs. 1 style transfers to new beneficiaries that are followed by larger ones, and add cooling periods and dual approval for new beneficiaries and settlements.

  6. 06

    Protect Passwords, Keys, and Crypto Wallets

    Remove passwords and keys hidden in code, use short-lived access tokens, and rotate credentials regularly. Store cryptocurrency wallet keys in secure hardware or offline storage and require multiple approvals for withdrawals.

  7. 07

    Control AI Tool Usage and Respond Decisively to Incidents

    Block unauthorized offensive AI tools on company systems and monitor how AI services and API keys are used. If an incident occurs, isolate affected systems, preserve logs, reset all credentials, and assume the attacker may still be present until cleanup is fully verified.


Potential MITRE ATT&CK TTPs

Reconnaissance
T1595Active Scanning
T1595.002Vulnerability Scanning
Resource Development
T1588Obtain Capabilities
T1588.002Tool
T1588.007Artificial Intelligence
Initial Access
T1190Exploit Public-Facing Application
T1078Valid Accounts
T1078.004Cloud Accounts
Execution
T1651Cloud Administration Command
T1648Serverless Execution
T1059Command and Scripting Interpreter
Persistence
T1505Server Software Component
T1505.003Web Shell
Credential Access
T1552Unsecured Credentials
T1552.001Credentials In Files
T1552.004Private Keys
Lateral Movement
T1550Use Alternate Authentication Material
T1550.001Application Access Token
Command and Control
T1071Application Layer Protocol
T1071.001Web Protocols
T1090Proxy
T1090.002External Proxy
Exfiltration
T1041Exfiltration Over C2 Channel
Impact
T1657Financial Theft

References

References

What Next?

At Hive Pro, it is our mission to detect the most likely threats to your organization and to help you prevent them from happening.

Reduce real exposure. Not just vulnerability volume.