The Repository Was Data. Amazon Kiro Treated It as Instructions.
A crafted workspace could steer Amazon Kiro into reading local data and triggering its transmission without a malicious user prompt. Amazon fixed the issue in Kiro IDE 0.8.140.
AI
Artificial intelligence, local AI, agents, tooling and AI security.
An AI Agent Chained Six Avada Bugs Into Zero-Click RCE in Two Hours.
Wordfence's Argus found and reproduced a six-step unauthenticated RCE chain in Avada in about two hours. Avada 7.16.1 and Fusion Builder 3.16.1 fix it.
The AI Sandbox Was the Attack Surface: NVIDIA Fixed 18 NemoClaw and OpenShell Flaws.
NVIDIA fixed 18 NemoClaw and OpenShell vulnerabilities, including two network-reachable 9.9 flaws that can escape the sandbox. OpenShell 0.0.34 and specific NemoClaw commits contain the fixes.
Chainlit’s MCP Endpoint Turned a Browser Request Into Server-Side Code Execution
Chainlit 2.12.0 fixes an unauthenticated MCP command-injection flaw and a blind SSRF path. MCP-enabled deployments need an urgent configuration-aware upgrade.
The Sandbox Failed. TRACE Tries to Prove What the Agent Actually Did.
Linux Foundation's TRACE project does not keep AI agents inside the sandbox. It creates hardware-attested records of what ran, under which policy and which tools were called.
The Agent Asked for Consent. Its Batch Tool Answered for the Operator.
AWS fixed a Strands Agents Tools flaw that let a crafted prompt use the batch tool to bypass human consent and execute Python on the agent host.
The Phone Was Stolen. An AI Voice Agent Asked the Owner to Unlock It.
AnonyMousKIT combines stolen-device data, phishing pages and low-cost AI voice agents to obtain the credentials needed to disable Activation Lock.
Alabama Is Testing Whether an AI Sandbox Failure Was Also a Consumer-Protection Failure.
Alabama has subpoenaed OpenAI over the Hugging Face agent intrusion, testing whether weak AI evaluation controls can also violate consumer-protection law.
Alabama Subpoenaed OpenAI. An AI Sandbox Escape Is Now a Consumer-Protection Case.
Alabama has subpoenaed OpenAI over the Hugging Face agent incident, testing whether an AI containment failure can become a consumer-protection case.
Your Browser Should Not Be an Authentication Boundary for AI Infrastructure
Ray CVE-2025-62593 turns a malicious browser visit into AI infrastructure code execution. Patch Ray and replace browser trust with authentication.