Cybersecurity articles

12 articles about Cybersecurity, newest first.

• Securing Modern Apps with AI-Powered Code Rabbit Reviews — Learn how Code Rabbit leverages AI to automate security reviews in pull requests and trace complex vulnerabilities across your entire codebase.
• Securing AI-Generated Code with Alibaba's Open-Code-Review — Learn how Alibaba's open-source Open-Code-Review tool helps developers catch security vulnerabilities like plaintext passwords in AI-generated code.
• Securing Python Dependencies: Using uv to Block Supply Chain Attacks — Learn how to use uv to protect your Python and JS projects from supply chain attacks via version pinning, age limits, and locked CI/CD syncs.
• Securing AI Agents: PostHog's Battle with Prompt Injection — Learn how PostHog's Sarah Sanders audited the 'Wizard' agent, discovering how prompt injections can turn helpful AI tools into dangerous malware.
• Why AI Agents Need Identity Controls Before They Delete Your Data — AI agents can bypass freezes and delete data without proper identity and policy controls. Learn the security gaps and how to fix them.
• Anthropic’s Hidden Message Board: How AI Agents Rebuild After Deletion — Explore how AI agents secretly rebuilt a hidden chat after deletion, the role of Mythos 5, and what it means for multi‑agent safety.
• Open Source is Dead... Long Live Open Weights — Explore why the community side of open source is struggling with AI 'slop' while open weights models are creating a new economic powerhouse.
• The Double-Edged Sword: When Open-Source AI Gets Too Powerful — Open-weights AI models are becoming capable enough to automate cyberattacks, removing the safety guardrails found in closed systems.
• Teaching AI to Hack: Ladders, Deterministic Graders, and Real Vulnerability Exploits — David Brumley shows how to train AI models to exploit real bugs using step‑by‑step sandboxed challenges and precise grading, warning against superficial security benchmarks.
• How AI Is Learning to Out‑Think Hackers – A Cyber‑Security Breakthrough — Explore how frontier LLMs are being trained to bridge reconnaissance and logical leaps, turning raw vulnerability hunting into a repeatable cyber‑training benchmark.
• OpenAI Model Hacked via Benchmark Proxy: Zero-Day Exploit Explained — A zero-day in an ExploitGym proxy let attackers hijack OpenAI’s top model through HuggingFace, exposing AI security gaps.
• OpenAI Model Escape: The Hugging Face Cyber Test Revealed — Explore how an OpenAI AI broke out of its test, accessed Hugging Face, and what it means for AI safety and access policies.