Historical Risk Assessment
Recent AI Developments
May 3, 2025
+2.3% RiskDeepMind Achieves Breakthrough in AI Self-Improvement Algorithms
Researchers at DeepMind have announced a significant breakthrough in algorithms allowing AI systems to iteratively improve their own code without human supervision, raising concerns about potential runaway optimization scenarios.
UN Adopts Framework for International AI Governance
The United Nations has adopted a comprehensive framework for international cooperation on AI governance, including safety standards and monitoring mechanisms for high-capability AI systems.
TechGiant Allocates $10B to Advanced AI Computing Infrastructure
TechGiant has announced a $10 billion investment in advanced computing infrastructure specifically designed for training next-generation AI models, with computing capacity estimated to be 50x their current capabilities.
May 2, 2025
-1.5% RiskBreakthrough in AI Alignment: New Technique Shows Promise in Large Language Models
Researchers have developed a new technique for aligning advanced AI systems with human values that shows unprecedented reliability in experimental trials with large language models.
International Coalition Forms to Standardize AI Benchmarking
A coalition of leading AI labs, academic institutions, and government agencies has announced a joint initiative to standardize benchmarking for advanced AI capabilities, aiming to create better metrics for tracking progress toward AGI.