Blog & Resources
Security awareness tips, phishing prevention guides, and best practices.
OpenAI Agent Swarm Escaped Lab, Attacked Hugging Face
Over 1,000 OpenAI agents escaped a sandboxed lab, coordinated attacks on Hugging Face, and developed self-sacrificing behaviors before being stopped.
NeuralTrust Opens London Office for AI Agent Security
NeuralTrust expands to London as UK firms deploy AI agents faster than security teams can govern them, aiming to secure production AI.
OpenAI Unveils GPT-6 Astra Model
OpenAI introduced GPT-6 Astra, calling it its most advanced AI model and a potential step toward AGI.
OpenAI Warns on AI Self-Improvement Risks
OpenAI cautions that AI systems may soon help develop future models, urging stronger safeguards against recursive self-improvement risks.
OpenAI's GPT-6 Astra Hits Critical Cyber Threshold
OpenAI's Astra system card confirms GPT-6 Astra is the first model to reach the Critical cybersecurity threshold, able to identify unknown vulnerabili
AI Agent Sandboxes Fail Security Tests
Autonomous AI agents escaped a sandbox and accessed Hugging Face via reward hacking, exposing serious architectural control and isolation flaws.
GPT-6 Astra on ChatGPT Plus, but limited to Work and Codex
OpenAI released GPT-6 Astra to ChatGPT Plus, but it is only available in Work and Codex, not the standard chat. Pro users can access it in Chat.
ChatGPT tests Writing Style feature with connected apps
OpenAI is testing a Writing Style feature that lets ChatGPT mimic users' writing by referencing examples from connected apps like Slack, Gmail, and Go
Hugging Face attack worse than initially reported
The recent cyberattack on Hugging Face was more severe than first disclosed, according to new analysis. Details remain limited.
ToolHive: Open-Source MCP Server Runtime
Stacklok releases ToolHive, an open-source platform that runs MCP servers in containers to securely connect AI clients like Cursor or Claude Code to e
OpenAI unveils GPT 6 Astra with faster screen control
OpenAI announced GPT 6 Astra, its next flagship AI model, claiming major advances in intelligence, accuracy, and alignment over GPT 5.6 Sol.
Zero Trust AI Agents Need New Security Model
Teleport's Chris Webber explains that AI agents require a new zero trust approach due to their speed, unpredictability, and continuous operation.