CMU Builds Self-Improving AI Agents via Constitutional RL
Carnegie Mellon researchers introduce Constitutional RL, a framework enabling AI agents to self-improve while following …
🏷️ AI Safety
136 articles about 'AI Safety'
AI Regulation Must Balance Innovation and Safety
Governments worldwide race to regulate AI, but striking the right balance between fostering innovation and protecting th…
Schmidt Warns AGI May Arrive Before 2028
Former Google CEO Eric Schmidt says artificial general intelligence could emerge within 3 years, raising urgent question…
Senate AI Panel Unveils Bipartisan Foundation Model Rules
A bipartisan Senate committee proposes a new regulatory framework targeting foundation model developers with transparenc…
OpenAI Advances Superalignment Research for AI
OpenAI publishes new research on superalignment techniques aimed at keeping frontier language models safe and aligned wi…
What Could Go Wrong With Trump's AI Safety Tests
Experts warn Trump's AI safety testing framework faces major pitfalls despite vindicating Biden's original approach.
Critical Jailbreak Flaw Hits All Major LLM Providers
Security researchers uncover a universal jailbreak vulnerability that bypasses safety guardrails across GPT-4, Claude, G…
Anthropic Unveils Constitutional AI 2.0 Research
Anthropic publishes new Constitutional AI 2.0 paper advancing scalable oversight methods for safer, more aligned AI syst…
OpenAI Whistleblowers Urge Senate Action on AI Safety
Former OpenAI employees testify before U.S. Senate, raising alarms about internal safety culture and calling for federal…
Trump Reverses Course on AI Safety Testing
After dismissing Biden's AI safety framework, the Trump administration now embraces testing requirements following alarm…
Kid With Fake Mustache Fools Age-Verification AI
Meta overhauls its age-verification system with AI that analyzes bone structure and height after simple disguises fooled…
Ex-OpenAI CTO Murati Says She Could Not Trust Altman
Mira Murati testified under oath that Sam Altman made false representations about safety review processes for a new AI m…