This is a special weekend edition examining the growing gap between AI adoption speed and AI security readiness. Four stories that reveal why your AI stack may be more exposed than you think — and what to do about it.

1. Critical Vulnerabilities in Amazon Bedrock, LangSmith, and SGLang: Security researchers discovered flaws that could expose training data, API keys, and allow remote code execution in widely-used AI platforms.
2. The Visibility Gap: A new report shows 67% of security leaders cannot see their own AI-related vulnerabilities, while AI-powered attacks have increased 300% year-over-year.
3. NVIDIA OpenShell: Released at GTC 2026, this open-source runtime sandboxes autonomous AI agents with kernel-level isolation and policy enforcement that the agent itself cannot over ...