EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Jul 31·5 min read
Google fixed more Chrome bugs in June than over the past two years, thanks to AI
TL;DR
In June, Google utilized AI to identify and fix more Chrome security bugs than in the past two years combined. This advancement highlights the transformative impact of Large Language Models in software security.
✦ Why It Matters
Consider integrating AI tools for vulnerability detection in your software development process today.
Key Takeaways
- Explore AI tools for automated vulnerability detection.
- Implement automated triaging to save developer time.
- Adopt memory-safe languages like Rust for new projects.
- Stay updated on Chrome's security release cadence.
How It Works
Google's approach involves using LLMs to enhance security fuzzing and automate the triaging process. By integrating AI models like Gemini and Big Sleep, they can efficiently identify vulnerabilities across the Chrome codebase.
The automated triage system filters, reproduces, enriches, and assigns bugs, significantly speeding up the process. Additionally, multi-agent workflows are employed for fixing vulnerabilities, where a fixing agent generates candidate fixes evaluated by a critic agent.