OpenAI has just launched Codex Security in research preview, a new agentic tool designed to transform how developers find and fix vulnerabilities. Here’s the breakdown of what this means for the security landscape: - Beyond Static Scanning: Unlike traditional scanners that flag every potential issue, Codex Security acts like a "security researcher." It builds a deep, project-specific threat model to understand the context and intent of your code. - Validated Findings: To cut through the noise, the agent pressure-tests its findings in a sandboxed environment. This means it only surfaces high-confidence vulnerabilities that are actually exploitable, reducing false positives by over 50%. - Actionable Fixes: It doesn't just find bugs; it proposes patches. These fixes are grounded in your system’s specific architecture, making them safer to review and merge without causing regressions. - Scaling Defense: In its first 30 days of testing, it scanned 1.2 million commits and identified over 11,000 high-severity and critical issues. - Support for Open Source: Through the "Codex for OSS" program, OpenAI is providing free access to maintainers of critical open-source projects to help secure the digital foundations we all rely on. - Availability: It's rolling out now to ChatGPT Pro, Enterprise, Business, and Edu customers.

Google is building toward AI that understands text, images, video, audio, and documents all in one system. That means it can connect meaning across different types of content instead of treating each one separately. 📊 Supports up to 6 images and 120 minutes of video or audio in one query ⚙️ Includes MRL for more efficient embeddings 🚀 A major step for multimodal search and retrieval This is where AI search is heading.