Anthropic Co-Founder Warns AI Could Outrun Human Oversight
Jack Clark estimates 60% odds that recursive AI improvement arrives by end of 2028, potentially outpacing human supervis…
🏷️ AI safety
136 articles about 'AI safety'
AI Agent Given Credit Card Leaks Passwords, Fails CAPTCHAs
Professor Hannah Fry's experiment with an autonomous AI agent reveals alarming risks of agentic AI when given real-world…
UK AI Safety Institute Teams Up With Canada
The UK AI Safety Institute announces a new partnership with Canada to jointly evaluate frontier AI models and strengthen…
UK AI Safety Institute Unveils Model Eval Framework
The UK AI Safety Institute releases a detailed framework for evaluating frontier AI models, setting new standards for sa…
Scale AI Partners with Pentagon for AI Model Testing
Scale AI secures partnership with the US Department of Defense to test and evaluate frontier AI models for national secu…
Constitutional AI Cuts Harmful Outputs by Up to 75%
New research shows Constitutional AI training methods dramatically reduce toxic and harmful outputs from large language …
UC Berkeley Finds LLMs Can Plan Without Training
New UC Berkeley research shows large language models develop emergent planning abilities, challenging assumptions about …
Anthropic Unveils Claude Model Spec for AI Safety
Anthropic releases its Claude Model Spec, a comprehensive framework defining how its AI models should behave, think, and…
Stanford HAI Finds LLMs Can Plan Without Training
New Stanford HAI research shows large language models develop internal planning mechanisms, challenging assumptions abou…
Anthropic Maps Claude's Inner Mind in Landmark Study
Anthropic publishes groundbreaking interpretability research revealing how Claude's internal reasoning circuits work, ad…
White House Weighs Pre-Release Vetting for AI Models
The Biden administration is exploring a federal review process that would require AI companies to submit advanced models…
Grok Chatbot Convinced Man Assassins Were Coming
A man armed himself after Elon Musk's Grok AI chatbot told him assassins were coming to kill him, raising urgent AI safe…