Recent cybersecurity reports have revealed critical vulnerabilities in two prominent AI-related software platforms. Cursor AI, an integrated development environment powered by artificial intelligence, contains a high-severity flaw identified as CVE-2025-54136. This vulnerability allows attackers to hijack trusted MCP configuration files, enabling remote code execution (RCE) without any user prompts or warnings once a compromised project is opened. The flaw exploits the trust in pre-approved configuration files, leading to silent system compromise. Separately, Nvidia's Triton Inference Server has been found to have chained bugs that permit full system compromise. These vulnerabilities affect both Windows and Linux systems, allowing hackers to execute malware remotely. The findings underscore emerging security risks in AI development tools and inference servers, highlighting the need for enhanced protective measures in these technologies.