The UK AI Safety Institute’s AI assurance pathway needs to enter a second phase

Written by

in

The UK AI Safety Institute (AISI) needs to expand its assurance remit while maintaining its edge in pre-deployment evaluations of frontier models, in order to support the domestic AI assurance ecosystem. Through the open-source evaluation framework Inspect, government procurement and AI verification research, AISI can reinforce its industry influence. However, it has no authority to compel companies to share models, and access was restored after 17 days, showing that the current model still faces risks.

Source:Tech Policy Press · Read original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *