RealFeedApp OpenAI found 1000 vulnerabilities in the Linux Kernel
The study demonstrates that using LLMs to analyze code enables finding vulnerabilities orders of magnitude cheaper and faster than traditional methods, though the process remains non-deterministic.
- Field
- computer security
- What they did
- Researchers evaluated how expensive and slow it is to detect software vulnerabilities using large language models, showing that the cost to analyze a single vulnerability can reach thousands of dollars.
- Why it matters
- This helps explain why only large corporations and venture-funded labs can afford such research, while other organizations find it financially inaccessible.
#llm#software security#vulnerability detection#openai#anthropic
