The AI Safety Summit and Labour's Response to National Security Risks from AI

EXECUTIVE SUMMARY

The UK is hosting the first AI Safety Summit on ‘frontier AI systems’ on 1-2 November at Bletchley Park. Frontier systems are highly-capable general-purpose AI systems, and the most advanced models will likely be released by leading AI labs next year. 

Frontier AI systems pose three types of risk: 

1. Societal Risks - arising from the integration of frontier AI into society; 
2. Misuse Risks - including the ability to help terrorists, strategic adversaries and rogue states develop biological, chemical or cyber weapons; and 
3. Loss of Control Risks - where we lose the ability to control AI tools, or build AI tools capable of evading attempts to shut them down.

The UK intelligence community (GCHQ and MI5) has issued multiple warnings over the potential national security risks of AI, as has the US Government. Companies themselves are also warning of the potential risks their technology could pose. Polling indicates that the UK public is significantly concerned about the risks of AI.

Labour will always put the security of the British people first. If Labour downplays the significance of AI national security threats - disregarding warnings from experts in security and AI and the positions of our close allies - it is vulnerable to potential attacks from the Conservatives that it is weak on security issues.  

Labour should take action to tackle the full range of potential challenges arising from AI - ranging from bias, discrimination and misinformation to safety and security concerns.