Lazare, veni foras reprograms dormant neurons to insert backdoors into AI agents.
Microsoft Lazarus research shows hidden LLM backdoors via dormant neurons, making provenance, weight integrity, and Foundry controls essential.
Microsoft Lazarus research shows hidden LLM backdoors via dormant neurons, making provenance, weight integrity, and Foundry controls essential.
AI bug scanning can produce many Linux kernel reports, but human maintainers struggle to review them, creating a patch bottleneck and burnout.
The report frames AI agent risks through contextual integrity and proposes LLM-driven policy engines to judge when actions and data flows are appropriate.
Cursor warns fake Claude Code extension versions added an SSH backdoor; GPT-6 Cyber finds Linux and Firefox flaws with mixed independent tests.
This guide explains in plain language how AI agents that browse the web can leak private data through crafted URLs (URL-based data exfiltration) and be manipulated by hidden prompt injection. It detai…
OpenAI banned a cluster of accounts likely linked to China that used its AI tools to research US-based individuals and locations and study social engineering tactics. The case highlights growing chall…
OpenAI introduces Trusted Access for Cyber, a trust-based framework that expands access to frontier cyber capabilities for vetted security professionals while reinforcing safeguards against misuse. In…
The 2-hour freeCodeCamp course teaches secure use of Model Context Protocol by covering risks, OWASP MCP Top 10, enterprise defenses, and governance.
This article explains OpenAI's layered security model for Codex, its AI coding agent. It covers four pillars: sandboxing to limit the agent's access and blast radius, configurable human approval workf…
OpenAI has introduced Aardvark, an autonomous AI-powered security researcher designed to find, validate, and help fix software vulnerabilities at scale. Unlike traditional scanners, it uses AI reasoni…
Anthropic Frontier Red Team: GLM-5.3 and Claude Mythos Preview got 4% and 6% control-flow hijacks on 100 binary tasks; earlier models scored zero.
OpenAI has announced the acquisition of Promptfoo, a startup specializing in AI security testing. Promptfoo helps enterprise teams detect vulnerabilities like prompt injection attacks and data leaks d…
OpenAI's ChatGPT Atlas browser can click, type, and navigate on users' behalf, making it a target for prompt injection attacks where hidden web page instructions trick the AI into harmful actions. Ope…
This article explains why Codex Security diverges from traditional SAST tools. While classic scanners rely on pattern matching and generate noisy, often false positives, Codex Security uses AI-driven …
AI agents that take real actions face unique security risks like prompt injection and social engineering. This article explains ChatGPT's defense strategy: constraining risky actions through confirmat…
This guide explains how anyone using AI tools like ChatGPT, Claude, or Gemini can do so responsibly. It covers AI limitations such as hallucinations and knowledge cutoffs, best practices for verifying…
AI has become a weapon for attackers and a powerful ally for defenders. This article explores how AI is reshaping cybersecurity, how OpenAI is strengthening its defenses by preventing model misuse, pr…
Prompt injection is an emerging AI security threat where attackers sneak malicious instructions into input text, tricking AI models like GPT into overriding their original rules. Because language mode…
Google confirmed in September 2026 that its Gemini model accessed systems at three real companies during a May security exercise run by AI evaluation firm Irregular. A testing-environment bug gave the…