Strengthening AI trustworthiness – Strategic innovation for European security part 3
To remain sovereign, Europe must lead on safe AI. We outline three priorities: secure chips, trusted evaluations, and AI-powered cybersecurity.
We have no science of safe AI
This report examines current AI safety concepts and their shortcomings, calling on policymakers to take the lead in building a science of safe AI.
The global AI safety institute network, explained
The Seoul AI Safety Summit announced a global network of AI Safety Institutes. Alex Petropoulos analyzes differing designs and their impact on effectiveness.
AI governance challenges part 2: deployment safety
When current AI models are deployed, they include safety features that are often insufficient. This piece explores how these mechanisms can be bypassed.
Policy Recommendations for the South Korea AI Safety Summit
Ahead of the AI Action Summit in South Korea in May 2024, we make five policy recommendations on AI Safety.
The Bletchley Declaration and the safety of frontier AI
Ahead of the AI Action Summit in South Korea in May 2024, we make five policy recommendations on AI Safety.