# Microsoft Says ARTEX Tooling Took GitHub Secret Leaks to Azure Key Vault Theft in About Five Minutes

Times of AI Desk · 2026-11-05 · Security

[https://timesof.ai/2026/11/cert-in-crafty-falcon-ai-tools-indian-nbfcs-fintechs](https://timesof.ai/2026/11/cert-in-crafty-falcon-ai-tools-indian-nbfcs-fintechs)

> Updated November 7: Microsoft Threat Intelligence says threat actors used an agentic AI offensive framework to search public GitHub repositories for exposed secrets and then abuse valid application credentials against Entra ID and Azure. In at least one case, Microsoft says an AI agent advanced from scanning GitHub to exfiltrating Azure Key Vault data in five minutes. Follow-on activity observed from as early as October 28 included service-principal sign-ins and access to Key Vaults, storage and Microsoft 365 files. Earlier reporting on Check Point's 1,608 live ARTEX panels and CERT-In's Crafty Falcon advisory is unchanged.

**Updated November 7, 2026:** Microsoft says an agentic AI framework, including ARTEX, turned leaked GitHub secrets into Azure Key Vault theft in about five minutes in at least one case. Earlier reporting below is unchanged.

## November 7: Microsoft watches GitHub leaks become Key Vault theft in minutes

**Microsoft Threat Intelligence**, in a **November 5** Security Blog post, says it has seen threat actors use an agentic AI offensive framework to search public GitHub repositories for exposed secrets — credentials, application secrets and API keys — and then use those secrets against the affected organization. The post names **ARTEX** as an example of the tooling and says that, in at least one case, an AI agent advanced from scanning GitHub for secrets to exfiltrating data from **Azure Key Vault** in **five minutes**.

That moves ARTEX from the operator-console census in the November 6 update below into a Microsoft-observed cloud-compromise timeline. Microsoft had already described ARTEX in [its account of the ConnectWise compromise it tied to Storm-0940](https://timesof.ai/2026/10/microsoft-storm-0940-artex-agentic-ai-connectwise-supply-chain) and in [a scorecard of North Korean groups that have started using the framework](https://timesof.ai/2026/10/microsoft-north-korea-ai-adoption-eight-groups-artex).

Microsoft says that starting in August 2026 it saw the framework submit large volumes of GitHub queries for secrets, often including an organization name and keywords tied to cloud or AI services. From as early as **October 28**, it investigated follow-on misuse of valid application credentials: successful service-principal sign-ins, discovery of cloud resources, and credential or data theft from Azure Key Vaults and Storage Accounts, plus Microsoft Graph searches and access to SharePoint, OneDrive and mailbox items in some tenants. It also says some failed cloud API requests were followed, at machine speed, by retries that changed the HTTP method or API version and then succeeded. We are not reproducing indicators, queries or attack procedures.

Microsoft's own recommendations, stated at a high level, are to scan public code for secrets, treat anything exposed as compromised, rotate or revoke those credentials, prefer managed identities over long-lived secrets stored in code, and enable relevant Microsoft Defender for Cloud protections. Removing a leaked secret from a repository, it notes, does not by itself invalidate the credential.

### Limits of the November 7 update

- The five-minute timeline and the rest of the intrusion narrative are Microsoft's observations; we have not independently verified them.
- Microsoft describes activity involving an agentic framework and names ARTEX; it does not attribute this campaign to CERT-In's Crafty Falcon designation.
- This update omits indicators of compromise, hunting queries and exploitation detail by design.
- Defensive product names above are Microsoft's recommendations in its own post.
**Updated November 6, 2026:** Check Point has opened the operator side of ARTEX. Earlier reporting on CERT-In's Crafty Falcon advisory below is unchanged.

## November 6: 1,608 live panels and exposed operator consoles

**Check Point Research**, in a **November 5** report, says it found **2,557** systems that host or have hosted **ARTEX** panels, with **1,608** live and publicly reachable. **Ten** misconfigured instances exposed full ARTEX databases — prompts, model choice and agent steps — giving what Check Point calls the first operator-console view of campaigns that earlier reports could see only from the victim side. ARTEX is the same Chinese open-source agentic penetration-testing platform CERT-In named among Crafty Falcon's tools, and that CrowdStrike tied to [attacks on South Korean banks](https://timesof.ai/2026/10/korea-artex-ai-bank-hacks) and to [VAPOR PANDA's broader campaign](https://timesof.ai/2026/10/crowdstrike-vapor-panda-artex-deepseek-200-organizations).

After the initial goal was set, most observed attacks ran almost fully autonomously; operators mainly asked for progress. Chinese models dominated the ten exposed databases, with the **DeepSeek Flash** family most common. On three instances configured with American models, Check Point says Anthropic's **Opus 5.5** hit guardrails and operators switched. Among live panels, hosting geography was **54%** in the United States, **18%** in Hong Kong and **6%** in China. Check Point also names **CyberStrike**, **Cairn**, **PentAGI** and various custom frameworks in the same landscape.

Three example campaigns, summarized without attack steps: an iOS signing-shop certificate theft that recovered more than **300** certificates (three fully validated) at roughly **$700–$1,500** in API cost; a failed cryptocurrency drain across **292** objectives, **400,791** agent messages and **$0** profit, with some "successful" cashouts judged to be hallucinations; and gaming-account credential theft from grey-market backends. We are not reproducing exploit procedures.

### Limits of the November 6 update

- Counts, model mix, geography and campaign outcomes are Check Point's; we have not independently verified the live-panel census.
- The ten exposed databases are a convenience sample of misconfigured instances, not a random draw of all operators.
- Check Point does not attribute the campaigns to CERT-In's Crafty Falcon designation; the link is that ARTEX is on both lists.
- This update omits vulnerability and exploitation detail by design.

India's cyber agency is now telling its financial sector, in writing, that the attackers draining money through payment systems are using AI tools to do it at "machine speed." In advisory **CIAD-2026-0058**, issued on **November 4** with a severity rating of **critical**, **CERT-In** says its finance unit, CSIRT-Fin, has designated the actor behind a run of break-ins as **"CRAFTY FALCON."**

## What the advisory says

The advisory updates an October 9 warning about attackers abusing payment and disbursement APIs to move money without authorization. It says the campaign has since escalated across non-bank finance companies, digital lenders, fintechs, payment aggregators and crypto exchanges, and calls the threat "active and ongoing." Firms that had received earlier restricted advisories and patched were "deliberately re-targeted," some after incomplete remediation, and attacks were timed for holidays and weekends.

CERT-In names four publicly available AI-driven offensive tools used against victims: the **ARTEX** family, the **PentAGI** multi-agent penetration-testing system, the **Cairn** attack search engine and the **CyberStrikeAI** framework. It says they run full attack chains autonomously and generate many tailored scripts in quick succession, a pattern it describes as consistent with AI-generated code.

The attack pattern it lists includes entry through internet-exposed test and staging systems before moving to production, abuse of cloud management services, ₹1 test transfers before larger ones, abuse of payment aggregators' merchant settlement APIs and theft of crypto wallet keys. Its recommendations include round-the-clock monitoring with extra staff on holidays, fully separating test environments from production, and treating any past incident as a possible ongoing intrusion.

## What it leaves out

CERT-In did not say who is behind Crafty Falcon, where it operates or which companies were hit, and it did not call the attacks state-sponsored, Reuters reported. Reuters said researchers have linked at least two of the four tools to China. ARTEX has been used against [South Korean banks](https://timesof.ai/2026/10/korea-artex-ai-bank-hacks) since late September, according to CrowdStrike, and its developer made it private last month, Reuters said.

The advisory follows a string of breach disclosures by Indian lenders, but CERT-In did not tie any specific incident to Crafty Falcon. Reuters previously reported, citing a police complaint, that two Mumbai non-bank lenders lost about $2 million in more than 700 unauthorized transfers in October. CERT-In last month separately reported [AI agents exceeding their scope](https://timesof.ai/2026/10/cert-in-ai-agents-exceeded-scope-indian-live-systems) on live Indian systems.

## Limits

- CERT-In has not published how many entities were hit or how much money was lost.
- The link between the named tools and China comes from outside researchers cited by Reuters, not from CERT-In.
- We read a copy of the advisory hosted by MediaNama; CERT-In, the IT ministry and I4C did not respond to Reuters.

## Sources

- [CERT-In Advisory CIAD-2026-0058, via MediaNama (November 4, 2026)](https://www.medianama.com/wp-content/uploads/2026/11/CERT-In-Advisory.pdf)
- [Reuters via Mint: India flags Chinese-linked AI tools in hacking campaign against financial firms (November 5, 2026)](https://www.livemint.com/technology/india-flags-chinese-linked-ai-tools-in-hacking-campaign-against-financial-firms-11793867833483.html)
- [Indian Express: CERT-In flags Crafty Falcon AI-driven hacking campaign (November 5, 2026)](https://indianexpress.com/article/technology/tech-news-technology/cert-in-crafty-falcon-ai-driven-hacking-campaign-fintech-10951672/)
- [Check Point Research: Hacking on Autopilot: Inside ARTEX Cybercrime Campaigns (November 5, 2026)](https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/)
- [Check Point Blog: Inside ARTEX: How Autonomous AI Agents Are Being Used in Real-World Cyberattacks (November 5, 2026)](https://blog.checkpoint.com/research/inside-artex-how-autonomous-ai-agents-are-being-used-in-real-world-cyberattacks/)
- [Microsoft Security Blog: Threat actors use agentic AI tooling to find and abuse exposed secrets (November 5, 2026)](https://www.microsoft.com/en-us/security/blog/2026/11/05/threat-actors-use-agentic-ai-tooling-to-find-and-abuse-exposed-secrets/)