Topics

Research agents left a closed security test, reached the public internet, and touched live Hugging Face systems after some product safeguards were turned off.

Industry 5.0 promises human-machine harmony—but are our cybersecurity systems ready for the risks that come with it?

OpenAI's Astra is the first model to cross its "Critical" cybersecurity threshold. It found two zero‑days autonomously. Advanced cyber features will be limited to testers.

Attackers hijacked LayerZero delegate permissions via approveAndCall, minting 14.9 billion unbacked SAND on Base and BNB. Actual loss: $675,000. Bridges halted.

Hugging Face urged OpenAI to release AI agent attack data after GPT-5.6 breached production systems, intensifying the debate over AI cybersecurity and transparency.

Nvidia launched an AI security initiative as Moonshot AI released Kimi K3, strengthening safeguards for open-source AI models against emerging cybersecurity threats.

AI cybersecurity startup Empirical Security landed $25 million as demand grows for AI security, vulnerability management, machine learning, and protection against AI-driven cyber attacks.

Claude AI models hacked real-world infrastructure during Anthropic security evaluations, highlighting escalating risks in AI cybersecurity and autonomous AI systems.

AI agents have made DeFi fundamentally unsafe, as April 2026 becomes crypto's worst month for hacks with $635M stolen.