TeamPCP ने Telnyx PyPI के ज़रिए WAV फ़ाइलों में स्टीलर छिपाया, LangChain & LangGraph की खामियाँ सीक्रेट्स को उजागर करती हैं, यूरोपीय आयोग AWS ब्रीच, डच पुलिस फ़िशिंग का शिकार, GitHub VS Code मालवेयर अभियान

TeamPCP ने PyPI पर Telnyx Python पैकेज को समझौता करके WAV ऑडियो फ़ाइलों में स्टेगनोग्राफी से क्रेडेंशियल-चोरी मालवेयर छिपाया। LangChain और LangGraph की तीन गंभीर कमज़ोरियाँ फ़ाइलसिस्टम डेटा, एनवायरनमेंट सीक्रेट्स और कन्वर्सेशन डेटाबेस को उजागर करती हैं। यूरोपीय आयोग अपने AWS वातावरण की ब्रीच की जाँच कर रहा है — कथित तौर पर 350 GB डेटा एक्सफ़िल्ट्रेट किया गया। डच पुलिस ने फ़िशिंग ब्रीच का खुलासा किया। GitHub पर नकली VS Code सुरक्षा अलर्ट डेवलपर्स को मालवेयर वितरित करते हैं।

1. TeamPCP Compromises Telnyx PyPI Package — WAV Audio Steganography Hides Credential Stealer

TeamPCP continues its unprecedented open-source supply chain campaign. On March 27, 2026, researchers from Aikido, Endor Labs, JFrog, Socket, and StepSecurity confirmed that the Telnyx Python package on PyPI had been compromised with two malicious versions. The attack represents a significant evolution in technique: credential-stealing payloads are now hidden inside WAV audio files using steganography.

Attack Chain Details

The malicious code is injected into telnyx/_client.py, triggering automatically when the package is imported. The three-stage attack chain works differently depending on the target OS:

• Windows: Downloads hangup.wav from C2, extracts a hidden executable, drops it into the Startup folder as msbuild.exe for persistence across reboots
• Linux/macOS: Fetches ringtone.wav from C2, extracts a collector script that runs in-memory, harvests credentials from environment variables, .env files, and shell histories
• All platforms: Exfiltrates stolen data as tpcp.tar.gz via HTTP POST to 83.142.209[.]203:8080
• Kubernetes: Abuses service account tokens to deploy privileged pods across all nodes for lateral movement

The entire chain operates within a self-destructing temporary directory, leaving near-zero forensic artefacts. This is the same WAV steganography technique TeamPCP previously used in the "kamikaze" wiper malware — now refined for credential theft.

How TeamPCP Got the Token

Endor Labs researchers believe the Telnyx PyPI token was harvested during the earlier LiteLLM compromise. TeamPCP's credential harvester swept environment variables, .env files, and shell histories from every system that imported LiteLLM. If any developer or CI pipeline had both LiteLLM installed and access to the Telnyx PyPI token, that token was already in TeamPCP's hands — creating a cascading supply chain compromise.

CRA and NIS2 Implications

• CRA Software Supply Chain: The cascading nature of this attack — from Trivy to LiteLLM to Telnyx — demonstrates exactly why the Cyber Resilience Act mandates vulnerability handling processes and coordinated disclosure across the entire dependency graph
• NIS2 Article 21(2)(d): Supply chain security must now account for credential exposure from one compromised dependency enabling attacks on unrelated packages
• SBOM urgency: Organisations without complete Software Bills of Materials cannot assess exposure to this cascading attack chain

What This Means for Your Organisation

• Immediately downgrade Telnyx to version 4.87.0 — the PyPI project is now quarantined
• Audit all systems that imported LiteLLM for exposed PyPI tokens, API keys, and credentials
• Check Kubernetes clusters for unauthorised privileged pods — TeamPCP's lateral movement component targets k8s
• Implement package integrity verification beyond simple version checks — steganographic payloads bypass traditional scanners

2. LangChain & LangGraph Vulnerabilities Expose Files, Secrets, and Databases

Cyera security researcher Vladimir Tokarev has disclosed three critical vulnerabilities in LangChain and LangGraph — the world's most popular frameworks for building LLM-powered applications. Each flaw provides an independent path to drain sensitive enterprise data:

The most severe flaw, CVE-2025-68664 (CVSS 9.3), allows attackers to leak API keys and environment secrets by passing a crafted data structure that LangChain interprets as a pre-serialised object rather than user input. This vulnerability — dubbed "LangGrinch" by Cyata who first identified it in December 2025 — can be exploited via prompt injection.

The Ripple Effect

Cyera warns that LangChain doesn't exist in isolation — it sits at the centre of a massive dependency web. Hundreds of libraries wrap, extend, or depend on LangChain. When a vulnerability exists in LangChain's core, it ripples through every downstream library, wrapper, and integration. With 52 million downloads per week for LangChain alone, the blast radius is enormous.

EU AI Act and CRA Context

• EU AI Act Article 15: AI systems must meet cybersecurity requirements — a CVSS 9.3 deserialization flaw in the world's most popular AI framework represents a systemic risk to high-risk AI systems built on LangChain
• CRA vulnerability handling: The gap between CVE-2025-68664's initial discovery (December 2025) and full patching raises questions about the CRA's "without delay" patching requirement
• GDPR Article 32: Conversation histories exposed via CVE-2025-67644 may contain personal data — organisations processing personal data through LangGraph are now at risk of GDPR enforcement

What This Means for Your Organisation

• Patch immediately: Update langchain-core to ≥1.2.22 and langgraph-checkpoint-sqlite to 3.0.1
• Audit all LangChain deployments for exposed API keys — rotate any keys that may have been accessible
• Review conversation histories stored in SQLite checkpoints for sensitive data exposure
• Assess all downstream dependencies that wrap LangChain — they may inherit these vulnerabilities

3. European Commission Investigates AWS Breach — 350 GB Allegedly Stolen

The European Commission is investigating what appears to be a significant breach of its Amazon Web Services (AWS) cloud environment. Key details reported by BleepingComputer:

• At least one AWS account was compromised
• The attack was quickly detected and the incident response team is investigating
• The threat actor claims to have stolen over 350 GB of data including multiple databases
• Screenshots provided as proof show access to employee information and an email server used by Commission staff
• The attacker states they will not extort the Commission but intend to leak the data publicly
• AWS confirmed their services were not compromised — the breach was on the Commission's side

This is the second Commission breach in two months. In February, the Commission disclosed that its mobile device management platform was hacked on January 30, in an incident linked to Ivanti EPMM code-injection vulnerabilities that also affected the Dutch Data Protection Authority and Finland's Valtori.

Regulatory and Political Implications

The irony is not lost on observers — the institution drafting Europe's cybersecurity laws is itself suffering repeated breaches:

• NIS2 compliance: As an EU institution, the Commission is subject to Regulation (EU) 2023/2841 on cybersecurity for Union institutions — repeated breaches raise questions about its own compliance posture
• Cloud security: The breach underscores risks of public cloud adoption by government institutions — the EU's own Cloud Code of Conduct and EUCS certification scheme are meant to address precisely this
• Data protection: If employee personal data was exfiltrated, the European Data Protection Supervisor (EDPS) will investigate under Regulation (EU) 2018/1725
• Credibility gap: The Commission proposed new cybersecurity legislation on January 20, 2026, to strengthen defences against state-backed actors — it must now demonstrate it can protect its own infrastructure

What This Means for Your Organisation

• Audit your AWS IAM configurations — compromised credentials remain the leading cause of cloud breaches
• Implement cloud security posture management (CSPM) with real-time alerting
• If your organisation interacts with the European Commission, monitor for data exposure from the alleged 350 GB leak
• Review your own cloud incident response — can you detect and contain a breach as quickly as the Commission claims?

4. Dutch Police Discloses Phishing Breach — Second Incident in 18 Months

The Dutch National Police has reported being targeted by a phishing attack that resulted in a security breach. According to the March 25 press release:

• The Security Operations Center (SOC) detected the incident quickly and immediately blocked access
• Impact appears limited — citizens' data and investigative information were not exposed or accessed
• A criminal investigation has been launched
• The police have not disclosed which systems or accounts were affected

In September 2024, the Dutch police suffered a more severe breach linked to a "state actor" that stole work-related contact information for multiple officers, including names, email addresses, phone numbers, and some private data. That investigation remains ongoing.

NIS2 and Law Enforcement Cybersecurity

• NIS2 scope: While law enforcement agencies are generally excluded from NIS2's scope, they are still subject to national cybersecurity frameworks — the Dutch NCSC provides guidance that the police are expected to follow
• Pattern of targeting: Two breaches in 18 months targeting the same organisation suggests persistent adversary interest — potentially from the same state actor behind the 2024 breach
• Post-breach improvements: After the 2024 breach, the police implemented stronger 2FA requirements and continuous monitoring — this time the SOC detected the attack quickly, suggesting those investments paid off

What This Means for Your Organisation

• Phishing remains the #1 initial access vector — even organisations with mature SOCs get phished
• Quick detection is as important as prevention — the Dutch police's SOC caught this before serious damage occurred
• After a breach, invest in detection capabilities alongside prevention — the improved 2FA and monitoring from the 2024 response helped contain this incident
• Conduct regular phishing simulations and measure time-to-detection, not just click rates

5. Fake VS Code Security Alerts on GitHub Spread Malware to Developers

Application security company Socket has uncovered a sophisticated, large-scale campaign targeting developers on GitHub. The attackers:

• Post fake security advisories in the Discussions section of thousands of GitHub repositories, with titles like "Severe Vulnerability — Immediate Update Required"
• Use realistic fake CVE IDs and urgent language to create a false sense of urgency
• Impersonate real code maintainers or researchers using newly created or low-activity accounts
• Post in an automated, coordinated fashion across thousands of repositories within minutes
• Trigger GitHub email notifications to all tagged users and project followers

The fake alerts include links to supposedly patched VS Code extensions hosted on Google Drive. Clicking these triggers a cookie-driven redirect chain leading to a JavaScript reconnaissance script that collects timezone, locale, user agent, OS details, and automation indicators before delivering a second-stage payload to validated victims.

Social Engineering at Scale

This campaign is notable for its abuse of trust signals:

• GitHub Discussions are a legitimate notification channel — developers expect security alerts here
• Email notifications arrive in inboxes alongside real GitHub alerts
• Google Drive links provide a veneer of legitimacy
• The campaign targets developers specifically — who are more likely to act quickly on security advisories

CRA Developer Security Obligations

• CRA secure development: The Cyber Resilience Act requires manufacturers to follow secure development practices — developers who fall for this campaign may unknowingly introduce compromised extensions into their development environments
• Supply chain integrity: If compromised developer machines have access to package registries or CI/CD pipelines, the blast radius extends well beyond the individual developer
• GitHub's responsibility: Under the DSA, GitHub as a hosting service has obligations around illegal content — automated spam impersonating security researchers may qualify

What This Means for Your Organisation

• Train developers to verify vulnerability reports through authoritative sources (NVD, CISA KEV, CVE.org) before acting
• Never download extensions or patches from external sources like Google Drive — use official marketplaces only
• Implement email filtering rules to flag GitHub notification emails containing external download links
• Audit developer workstations for VS Code extensions installed from non-marketplace sources

Today's Threat Landscape Summary