OpenAI has released a report on the progress of its Cybersecurity Grant Programme, launched in 2023 ,to support research at the intersection of cybersecurity and artificial intelligence. The program has received over 600 applications, demonstrating strong interest from the cybersecurity community.
Highlighted projects include:
1. UC Berkeley's Wagner Lab: Developing defenses against prompt-injection attacks in large language models (LLMs).
2. Coguard: Using AI to reduce software misconfiguration.
3. Mithril Security: Creating tools to secure LLM inference infrastructure.
4. Gabriel Bernadett-Shapiro: Developing AI training resources for security professionals.
5. Breuer Lab at Dartmouth: Working on new defense techniques against adversarial attacks on neural networks.
6. Security Lab Boston University (SeclaBU): Improving LLMs' ability to detect and fix code vulnerabilities.
7. CY-PHY Security Lab at UCSC: Exploring autonomous cyber defense agents using foundation models.
8. MIT CSAIL: Automating decision processes for red-teaming and exploring LLM capabilities in Capture-the-Flag challenges.
OpenAI is also expanding access to its AI tools for cybersecurity professionals, offering free ChatGPT Plus accounts and extending this to ChatGPT Team and Enterprise, starting with partners at the Research and Education Network for Uganda (RENU).