The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added CVE-2026-33017 to its Known Exploited Vulnerabilities catalog, warning that the critical code injection flaw in the Langflow AI development framework is being actively exploited to allow for unauthenticated remote code execution.
Vulnerability Details
Affected Product: Langflow versions 1.8.1 and earlier
Identifier: CVE-2026-33017
CVSS Score: 9.3 (Critical)
Exploitation Status: Actively Exploited
Risk & Impact
Triage: Immediate action is required due to confirmed active exploitation. Federal agencies are required to patch by April 8, 2026.
Attack Vector: An unauthenticated attacker can send a single crafted HTTP request to a public API endpoint (`POST /api/v1/build_public_tmp/{flow_id}/flow`). This allows attacker-controlled code to be passed directly to a Python `exec()` function without sandboxing, resulting in remote code execution.
Ease of Exploit: Trivial. Attackers began automated scanning and exploitation within 20-21 hours of the public advisory, even without a public proof-of-concept.
Action Plan
Immediate Action: Upgrade to Langflow version 1.9.0 or later, which fully addresses the vulnerability.
Workaround: If immediate patching is not possible, do not expose Langflow instances directly to the internet. Disable or restrict network access to the vulnerable endpoint (`/api/v1/build_public_tmp/{flow_id}/flow`).
Detection: Monitor for suspicious outbound traffic. Rotate all API keys, database credentials, and cloud secrets if any suspicious activity is detected.
Relevant professional terms
CISA (Cybersecurity and Infrastructure Security Agency)
A U.S. federal agency responsible for leading national efforts to understand, manage, and reduce risks to cyber and physical infrastructure. It maintains the Known Exploited Vulnerabilities (KEV) catalog to track security flaws actively used by malicious actors.
AI Agent Framework
A software platform with pre-built components, tools, and libraries designed to simplify the construction, deployment, and management of autonomous AI agents and their workflows. Langflow is an example of such a framework.
A critical vulnerability in Oracle WebLogic Server, identified as CVE-2026-21962, is being actively exploited.
According to a CloudSEK honeypot study, attackers began weaponizing the remote code execution (RCE) flaw on the same day that exploit code was publicly released.
Vulnerability Details
Affected Product: Oracle WebLogic Server (latest patched versions required)
Identifier: CVE-2026-21962
CVSS Score: 10.0 (Critical)
Exploitation Status: Actively Exploited / Public PoC Available
Risk & Impact
Triage: Immediate patching is required due to confirmed rapid exploitation starting the same day the exploit was published.
Attack Vector: An unauthenticated attacker can achieve remote code execution through automated scanning and probing of exposed servers (tools such as libredtail-http and Nmap Scripting Engine observed).
Ease of Exploit: High. Exploitation attempts originated from rented VPS infrastructure and were detected within hours of public exploit release.
Action Plan
Immediate Action: Apply the latest Oracle Critical Patch Update immediately.
Workaround: If patching is not immediately possible, restrict administrative console access from the internet, disable unnecessary protocols/ports, and deploy a web application firewall (WAF) with strict rules.
Detection: Monitor logs for suspicious scanning/probing activity and implement layered defenses including access controls for the administrative console.
Relevant professional terms
RCE (Remote Code Execution)
A type of vulnerability that allows a malicious actor to execute arbitrary code on a target machine across a network, potentially leading to a full system compromise.
Honeypot
A security mechanism that sets up a decoy computer system to attract and detect cyberattackers, allowing for the study of their methods and activities.
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added CVE-2026-33017 to its Known Exploited Vulnerabilities catalog due to evidence of active exploitation.
This critical code injection vulnerability in the Langflow AI framework allows for unauthenticated remote code execution and was reportedly exploited within 20 hours of its public disclosure.
Vulnerability Details
Affected Product: Langflow versions prior to 1.9.0.
Identifier: CVE-2026-33017
CVSS Score: 9.3 (Critical).
Exploitation Status: Actively Exploited.
Risk & Impact
Triage: Immediate patching is required; CISA has set a remediation deadline of April 8, 2026, for federal agencies.
Attack Vector: An unauthenticated, remote attacker can send a specially crafted HTTP POST request to the `/api/v1/build_public_tmp/{flow_id}/flow` endpoint. This allows them to supply arbitrary Python code that is executed by the server without sandboxing, leading to complete system compromise.
Ease of Exploit: Trivial. The attack requires no authentication and only a single HTTP request, making it easy to automate for mass scanning and exploitation.
Action Plan
Immediate Action: Upgrade all Langflow instances to version 1.9.0 or later.
Workaround: If immediate patching is not possible, restrict network access to the Langflow instance using firewall rules or a reverse proxy with authentication. Consider disabling the vulnerable endpoint if feasible.
Detection: Monitor for outbound connections to unusual ports or known callback services. Review web server logs for suspicious POST requests to the `/api/v1/build_public_tmp/` endpoint and look for unexpected process execution from the Langflow service.
Relevant professional terms
Remote Code Execution (RCE)
A type of vulnerability that allows a malicious actor to execute arbitrary commands or code on a target machine over a network, often leading to full system compromise.
Supply Chain Compromise
An attack that targets less-secure elements within a software development or distribution process, such as open-source libraries or build tools, to inject malicious code into a final product.
The pro-Ukrainian hacktivist group "Bearlyfy" has launched over 70 cyberattacks against Russian companies since January 2025, deploying a custom ransomware strain known as GenieLocker. The group's motives appear to be a mix of financial gain and political disruption.
Key TTPs
Initial Access: Exploiting vulnerabilities in external-facing services and applications.
Execution: Deployment of remote access tools to facilitate data encryption and system modification.
Campaign Analysis
Bearlyfy has evolved from a group with modest capabilities using leaked ransomware builders to a more sophisticated actor developing its own malware. This development of the custom GenieLocker ransomware signals a significant increase in the group's technical maturity and operational structure.
Targeting & Infrastructure
Target Profile: Initially small businesses, but now shifting to larger Russian corporate enterprises.
Infrastructure: The group has shown potential collaboration or shared infrastructure with other pro-Ukrainian hacking entities like PhantomCore and Head Mare.
Relevant Terms
Ransomware: Malicious software that encrypts a victim's files and demands a payment (ransom) for the decryption key.
Initial Access: The first stage of a cyberattack, where an adversary successfully gets a foothold inside a network.
A China-linked threat actor, Red Menshen, is conducting a long-term espionage campaign by embedding the stealthy BPFDoor implant within global telecommunication networks.
This strategic infiltration allows the actor to maintain persistent access to critical infrastructure for spying on government communications.
Key TTPs
Initial Access: Compromising internet-facing network appliances like VPNs, firewalls, and routers to gain an initial foothold.
Execution: The BPFDoor implant is activated by a specific "magic packet" sent by the threat actor, which triggers a reverse shell.
Defense Evasion: The malware operates at the kernel level, does not open any new ports, and can rename its own process, making it highly evasive to traditional security tools.
Campaign Analysis
This campaign represents a shift toward deeper, more persistent attacks on core infrastructure. By embedding "sleeper cells" within telecom providers, the actor ensures long-term access for high-level espionage, a tactic that is significantly harder to detect than traditional malware.
Targeting & Infrastructure
Target Profile: Telecommunication providers, government, finance, and logistics sectors, primarily in the Middle East and Asia.
Infrastructure: The actor utilizes compromised routers and Virtual Private Servers (VPSs) to relay commands to the BPFDoor implants.
Relevant Terms
BPFDoor: A stealthy backdoor for Linux systems that uses Berkeley Packet Filter (BPF) to passively monitor network traffic for a special command, avoiding detection by not opening ports.
Reverse Shell: A session where a compromised machine initiates a connection back to an attacker's machine, bypassing firewall rules that typically block incoming connections.
The threat actor group TeamPCP compromised the official 'telnyx' Python SDK on the PyPI repository, distributing malware to developers.
This attack is part of a broader, weeks-long supply chain campaign targeting developers and their credentials.
Key TTPs
Initial Access: Publishing malicious versions (4.87.1 and 4.87.2) of a legitimate package to the public PyPI repository.
Execution: Malicious code executes automatically and silently as soon as the package is imported by a developer's application.
Defense Evasion: On Windows, the payload is hidden inside a `.wav` audio file using steganography to bypass network inspection.
Campaign Analysis
This incident is the latest in an escalating campaign where TeamPCP compromises one developer tool to steal credentials and then uses that access to poison the next target in the software supply chain.
The group demonstrates increasing sophistication by using novel techniques like steganography and even pushing minor bug fixes for their own malware.
Targeting & Infrastructure
Target Profile: Developers and organizations utilizing the Telnyx Python SDK for AI and voice services.
Infrastructure: The Python Package Index (PyPI) is used for distribution, with a command-and-control (C2) server for payload delivery.
Actionable Intelligence
IPs:83.142.209[.]203:8080
Relevant Terms
Supply Chain Compromise: An attack strategy that targets a trusted third-party vendor or software to gain indirect access to the final target's systems.
PyPI (Python Package Index): The official third-party software repository for the Python programming language, used by developers to share and download code libraries.
DarkSword is a sophisticated JavaScript-based toolkit that enables attackers to compromise iPhones by exploiting multiple vulnerabilities.
The leak of this tool on public forums dramatically lowers the barrier for entry for less-skilled actors to conduct advanced attacks against vulnerable iOS devices.
Key Features
Zero-Interaction Infection: Compromises devices through watering hole attacks; users only need to visit a compromised website.
Full-Chain Exploit: Utilizes a chain of vulnerabilities in Safari's WebKit and the iOS kernel to achieve remote code execution and privilege escalation.
Extensive Data Exfiltration: Designed to steal a wide range of data, including credentials, messages, photos, and cryptocurrency wallet information.
Use Case (The "So What?")
The public release of DarkSword equips Red Teams with a powerful, ready-to-deploy tool for simulating advanced persistent threats against iOS environments.
For Blue Teams and SOCs, this leak signifies a critical threat, as a formerly state-level capability is now in the hands of common cybercriminals, requiring an urgent reassessment of mobile device monitoring and incident response playbooks.
Availability
A version of the DarkSword exploit kit has been leaked and is publicly available on GitHub.
Relevant Terms
Exploit Chain: A series of vulnerabilities chained together, where each exploit "unlocks" the next level of access required to ultimately compromise a system.
Watering Hole Attack: A strategy where an attacker compromises a website likely to be visited by a specific target group, rather than attacking the targets directly.
