On February 14, 2025, the UK Agency for Science, Innovation and Technology announced its transition to the UK AI Safety Institute’s UK AI Security Institute. Read the AI Now statement on migration below.
The partnership with AISI with the Defense Science and Technology Institute, the Ministry of Defense’s science and technology organization, is focusing on the UK government’s use of frontier AI within defense and national security equipment. This is in the lengthy part of the recent announcement that major AI companies will integrate frontier AI models into national security use cases. As our research demonstrated, these systems pose serious risks, including threats to national security, given the cyber vulnerabilities inherent in frontier AI models, which could be trained. Delicate data includes being able to be extracted by the enemy.
We welcome AISI signals and could investigate these risks as the dynamics of “AI races” have been increased, but applying fragmentary or superficial scrutiny under the banner of security, these Warns against approaches that give a clean gaze before the system is ready. These issues cannot be easily fixed or patched, and require a critical assessment of independent safety that needs to be isolated from industry partnerships. If our leaders are moving forward with plans to implement Frontier AI for defence use, they risk damaging our national security. This is a trade-off that AI benefits cannot be justified.
This statement may be attributed to the chief of AI Now Now Ai scientist Heidy Khlaaf.
