Image: SubstackThis Week in NLP #400 - by Robert Dale - This Week in NLP
• OpenAI has developed GPT-Red, a specialized AI model designed to red-team other LLMs to identify vulnerabilities and strengthen defenses against cyberattacks. • Researchers from Tracebit discovered that embedding prompt injections with stored secrets on AWS can effectively prevent AI hacking agents from compromising accounts. • The White House launched Gold Eagle, a centralized cybersecurity clearinghouse aimed at coordinating software vulnerability discovery and remediation for federal agencies and critical infrastructure.
thisweekinnlp.substack.com






