Category
Articles tagged under security on Open-TechStack.
A reported unauthorized access to Anthropic's restricted Mythos preview on April 21, 2026 is less about one leak than the vendor, contractor, and endpoint controls around frontier AI rollouts.
On April 7, 2026, Anthropic launched Project Glasswing to give selected defenders early access to Claude Mythos Preview. The real story is not the model demo. It is the new operating model for finding and fixing vulnerabilities before attackers catch up.
DarkSword reporting is a reminder that zero-click mobile spyware requires fast patching, evidence preservation, and disciplined incident response rather than panic or amateur forensics.
Microsoft’s April 6, 2026 research on AI-enabled device code phishing is not just another phishing report. It shows how generative AI, real-time automation, and trusted cloud infrastructure are combining into a faster account-takeover workflow.
On April 10, 2026, OpenAI said a compromised Axios package touched a GitHub Actions workflow used in its macOS app-signing process. Here is what happened, what did not happen, and why builders should care.
A practical 2026 security checklist for handling AI agent browser traffic, verified bots, signed requests, rate limits, forms, login surfaces, and CDN/WAF rules.
OpenAI now supports ChatGPT workspace agents for eligible Enterprise EKM workspaces. Here is the governance checklist before teams connect Slack, MCP servers, files, schedules, and apps.
Anthropic added self-hosted sandboxes and MCP tunnels to Claude Managed Agents. Here is the practical rollout checklist for execution boundaries, private MCP access, and agent operations.