The National Institute of Standards and Technology‘s AI Safety Institute should work to support organizations seeking to ensure the security of artificial intelligence systems by developing additional technical red-teaming guidance for dual-use foundation models.
Global technology trade association ITI said Thursday it had made this recommendation in response to the AISI’s consultation regarding a draft document it had released on the management of risks to dual-use foundation models, a type of AI technology underpinning a variety of common applications like internet search but also possessing the potential for advanced benefits like accelerated research.
ITI also proposed that the guidance tackle the roles, responsibilities and capabilities of various actors in the AI value chain and identify where responsibility-taking can occur.
Concerning transparency and disclosure, the guidance should also list the information to be disclosed by organizations and to whom they should be disclosed.
According to ITI Vice President of Policy Courtney Lang, having “a consistent understanding of misuse risks and ways to address them” is key to AI safety.
“By incorporating the tech industry’s feedback, NIST can strengthen its guidance document and provide a playbook for stakeholders, ensuring consistency, bolstering accountability, and mitigating risks for consumers and businesses,” Lang added.
