The Rapid Rise of Generative AI: Assessing risks to safety and security

Janjeva, Ardi; Harris, Alexander; Mercer, Sarah; Kasprzyk, Alexander; Gausen, Anna

Abstract

This CETaS Research Report presents the findings from a major project exploring the implications of generative AI for national security. It is based on extensive engagement with more than 50 experts across government, academia, industry, and civil society, and represents the most comprehensive UK-based study to date on the national security implications of generative AI. The research found that generative AI could significantly amplify a range of digital, physical and political security risks. With the rapid proliferation of generative AI tools across the economy, the national security community needs to shift its mindset to account for all the unintentional or incidental ways in which generative AI can pose security risks, in addition to intentionally malicious uses. The report provides recommendations to effectively mitigate the security risks posed by generative AI, calling for a new multi-layered, socio-technical approach to system evaluation.

Citation

Janjeva, A., Harris, A., Mercer, S., Kasprzyk, A., & Gausen, A. (2023). The Rapid Rise of Generative AI: Assessing risks to safety and security. Alan Turing Institute

• Report Type: Research Report
• Acceptance Date: Dec 15, 2023
• Publication Date: Dec 15, 2023
• Deposit Date: Mar 29, 2024
• Publicly Available Date: Apr 9, 2024
• Public URL: https://nottingham-repository.worktribe.com/output/33026957
• Publisher URL: https://cetas.turing.ac.uk/publications/rapid-rise-generative-ai#:~:text=Currently%2C%20generative%20AI%20tools%20are,satisfy%20accountability%20and%20oversight%20requirements.