Week of Apr 15–21, 2024

Wallace and co-authors at OpenAI argue that models treat system prompts and untrusted inputs with equal priority and propose an explicit instruction hierarchy that tells the model which instructions to follow when they conflict. Applied to GPT-3.5, they report large robustness gains against attack types not seen in training with minimal capability loss.

The NSA's Artificial Intelligence Security Center led joint guidance with CISA, the FBI and the national cyber centres of Australia, Canada, New Zealand and the UK on deploying and operating externally developed AI systems. It sets objectives to improve the confidentiality, integrity and availability of AI systems and to mitigate known vulnerabilities, organized around protecting, detecting malicious activity against, and responding to incidents involving AI systems.