Skip to main content
← All Tags

AI Security

36 articles in this category (Page 2 of 2)

AI NewsAI SecuritySoftware Architecture

Architecting Unexploitable AI Agents: Beyond Prompt Engineering

Learn how to reduce AI agent attack surfaces from 50% to 0% using five layered defense patterns instead of relying on system prompts.

Read more
AI NewsAI SecurityDevOps

GO-GATE: Implementing Two-Phase Commit Safety for Autonomous AI Agents

GO-GATE introduces database-grade Two-Phase Commit guarantees to AI agents, preventing runaway cloud bills and unsafe operations through risk-tiered execution.

Read more
AI NewsCybersecurityAI Security

Securing Agentic AI: From MCPs and Tool Access to Shadow API Key Sprawl

AI agents now automate software tasks, and a recent flaw (CVE-2025-6514) in an OAuth proxy impacted 500,000 developers, highlighting the risks of compromised Machine Control Protocols.

Read more
AI NewsCybersecurityAI Security

2 Separate Campaigns Probe Corporate LLMs for Secrets

Nearly 100,000 sessions targeted public LLM endpoints to map an expanding attack surface and potentially steal sensitive data.

Read more
AI NewsAI SecurityChatGPT

Continuously hardening ChatGPT Atlas against prompt injection attacks

OpenAI is proactively discovering and patching real-world agent exploits in ChatGPT Atlas using automated red teaming powered by reinforcement learning.

Read more
AI NewsAI SecurityMachine Learning

Picklescan Bugs Allow Malicious PyTorch Models to Evade Scans and Execute Code

Three critical Picklescan vulnerabilities (CVSS 9.3) enable malicious PyTorch models to bypass scans and execute arbitrary code.

Read more
AI NewsKubernetesAI Security

Security Holes in MCP Servers and How To Plug Them

Model Context Protocol (MCP) turns one year old with persistent security risks, but agentgateway offers a fix.

Read more
AI NewsCybersecurityAI Security

Balanced SOC Investment Cuts False Positives by 90% in Phishing Defense

A 2025 case study shows SOCs prevent sophisticated phishing attacks missed by detection tools, reducing false positives by 90%.

Read more
AI NewsCybersecurityAI Security

ServiceNow AI Agents Can Be Tricked Into Acting Against Each Other via Second-Order Prompts

Second-order prompt injection exploits ServiceNow agent discovery, enabling unauthorized data access and privilege escalation.

Read more
AI NewsAI SecurityMachine Learning

Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents

Anthropic's study reveals 250 malicious documents can create LLM backdoors, challenging scaling assumptions.

Read more
AI NewsCybersecurityAI Security

Google’s AI ‘Big Sleep’ Finds 5 New Vulnerabilities in Apple’s Safari WebKit

Google’s Big Sleep AI uncovered five critical Safari WebKit flaws, patched by Apple in November 2025.

Read more
AI NewsAI SecurityCode Agents

RedCodeAgent improves red-team evaluation of code agent security risks

RedCodeAgent uncovers 82 unique vulnerabilities in code agents, outperforming static benchmarks.

Read more