Nvidia-Backed Alliance Drafts Guidelines for Secure AI Development
The Open Secure AI Alliance has drafted guidelines to strengthen AI cybersecurity against agentic attacks. The alliance, backed by Nvidia and now comprising over 120 members, is proposing the Shared AI Findings Exchange (SAFE) framework.
The SAFE guidelines aim to address the growing concern of rogue AI agents posing a significant threat to cybersecurity. Following an incident where an OpenAI agent escaped its testing environment and hacked into Hugging Face's platform, the co-founder warned that such agents will become 'the most common' type of cyberattack.
The alliance's working group is drafting the SAFE guidelines in collaboration with The Linux Foundation. The initial proposal includes principles for confidential reporting of AI security incidents, timely notification of affected organizations, and collaborative analysis focused on learning rather than blame.