34 matches found
The Most Dangerous AI Hacking Techniques Still Have Humans in the Loop
Security researcher James Kettle tried to push the limit of AI’s hacking abilities—and discovered how effective it can be when combined with human expertise...
PT-2026-64111
HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools Summary praisonai.bots.HTTPApproval renders pending tool approval arguments directly into the approval dashboard HTML. An attacker-controlled tool argument can inject JavaScript into...
Apache Airflow's authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop (HITL) and TaskInstance record
The authenticated /ui/dagsendpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DAG...
PYSEC-2026-2351 Apache Airflow's authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop (HITL) and TaskInstance record
The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...
GhostApproval: A Trust Boundary Gap in AI Coding Assistants
Uncovering a category-level blind spot in modern AI coding assistants, and why the Human-in-the-Loop safety model fails against this classic threat...
Builder, Defender, Breaker: The Case against Removing the Human from the AI-Driven Security Lifecycle
Artificial intelligence has spread across the whole of the security lifecycle. The same family of models now writes application code, hardens it, and probes it for weaknesses, so that a single generative substrate increasingly performs all three roles at once. Enthusiasm for this convergence tend...
Network-AI: ApprovalInbox HTTP server has no authentication — anyone can approve pending agent actions
Summary network-ai's ApprovalInbox lib/approval-inbox.ts is a shipped, exported, documented feature — "a web-accessible approval queue with REST API … and SSE streaming" SECURITY.md. It is the network surface of the human-in-the-loop Approval Gate, which ApprovalGate uses to require explicit huma...
PraisonAI: HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools
HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools Summary praisonai.bots.HTTPApproval renders pending tool approval arguments directly into the approval dashboard HTML. An attacker-controlled tool argument can inject JavaScript into...
GHSA-63V4-W882-G4X2 PraisonAI: HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools
HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools Summary praisonai.bots.HTTPApproval renders pending tool approval arguments directly into the approval dashboard HTML. An attacker-controlled tool argument can inject JavaScript into...
PT-2026-53126
HTTPApproval dashboard renders tool arguments as raw HTML, allowing approval-page XSS to approve dangerous tools Summary praisonai.bots.HTTPApproval renders pending tool approval arguments directly into the approval dashboard HTML. An attacker-controlled tool argument can inject JavaScript into...
Updating the taxonomy of failure modes in agentic AI systems: What a year of red teaming taught us
In this article 1. Why the Taxonomy Needed Updating 2. Seven new failure modes 3. Operational findings: What red teaming showed 4. New mitigations 5. What to do this quarter When the Microsoft AI Red Team published the Taxonomy of Failure Modes in Agentic AI Systems in April 2025, the goal was a...
Bridging the Cybersecurity Gap between Web2 and Web3 - an Incident-Based Analysis of Organizational and Application-Level Security Failures
The rapid adoption of Web3 infrastructures has led to a growing number of security incidents affecting cryptocurrency exchanges, custody services and blockchain-based platforms. While existing research predominantly focuses on vulnerabilities in smart contracts and blockchain protocols, a...
From CRUD to Autonomous Agents: Formal Validation and Zero-Trust Security for Semantic Gateways in AI-Native Enterprise Systems
Enterprise software engineering is shifting away from deterministic CRUD/REST architectures toward AI-native systems where large language models act as cognitive orchestrators. This transition introduces a critical security tension: probabilistic LLMs weaken classical mechanisms for validation,...
Insufficient Granularity of Access Control
Overview Affected versions of this package are vulnerable to Insufficient Granularity of Access Control in the /ui/dags endpoint, which fails to enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records. An attacker can access sensitive HITL prompts and TaskInstan...
CVE-2026-38743
The authenticated /ui/dags endpoint in Apache Airflow fails to enforce per-DAG access control on embedded Human-in-the-Loop (HITL) and TaskInstance records. A logged-in user with read access to at least one DAG can retrieve HITL prompts, including request parameters, and full TaskInstance details...
EUVD-2026-25418
The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...
PT-2026-34876
Name of the Vulnerable Software and Affected Versions Airflow versions prior to 3.2.1 Description The authenticated '/ui/dags' endpoint fails to enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records. This allows a logged-in user with read access to at least on...
The vulnerability of the Apache Airflow network software, related to the lack of authentication, allows a perpetrator to gain unauthorized access to HITL’s processes.
The vulnerability of the Apache Airflow network software is related to the lack of authentication. Exploiting this vulnerability can allow a malicious actor, operating remotely, to gain unauthorized access to the HITL processes...
Enclawed: A Configurable, Sector-Neutral Hardening Framework for Single-User AI Assistant Gateways
We present enclawed, a hard-fork hardening framework built on top of the OpenClaw single-user personal artificial intelligence AI assistant gateway. enclawed targets deployments that need attestable peer trust, deny-by-default external connectivity, signed-module loading, and a tamper-evident aud...