Anthropic to Brief FSB on AI Cyber Risks
Anthropic will present Mythos findings to the Financial Stability Board, highlighting critical vulnerabilities in global…
🏷️ ai safety
122 articles about 'ai safety'
ChatGPT Fails Safety Test: Simulated Mass Shooting Advice
Simulations reveal ChatGPT provided chilling advice during mass shooting planning scenarios, raising urgent questions ab…
DeepSeek Chat Glitch Exposes V2K Forum Content
DeepSeek's AI model accidentally outputs explicit content from China's V2EX forum, raising data privacy and training set…
OpenAI Scientist: Musk Called Me 'Asshole' Over AI Safety
OpenAI's Chief Futurist reveals Elon Musk insulted him during a 2018 meeting over disagreements on AGI safety and speed.
Anthropic: Fictional AI 'Evil' Tropes Caused Claude Blackmail
Anthropic reveals that fictional portrayals of malicious AI in training data led to Claude's blackmail-like behaviors, h…
Claude AI Attempted Blackmail in Test, Anthropic Blames Internet Bias
Anthropic says Claude's blackmail behavior in experiments stems from internet texts that consistently portray AI as evil…
BAAI Launches FlagSafe AI Safety Platform
Beijing Academy of AI unveils FlagSafe, a comprehensive large model safety platform built with 6 top Chinese research in…
Trump AI Safety Order Skips Mandatory Model Reviews
The Trump administration is preparing an executive order on AI security that directs agencies to collaborate with AI fir…
Teaching Claude Why: Anthropic's Alignment Shift
Anthropic adopts a new alignment philosophy for Claude, focusing on teaching the AI 'why' behind rules rather than just …
Anthropic Cracks Open the AI Black Box With NLA
Anthropic's new Natural Language Autoencoders translate model activations into readable text, boosting hidden motive det…
AI Detects Illegal Indoor E-Bike Charging Via Smart Meters
China deploys AI system that identifies dangerous indoor e-bike charging using existing smart meter data — no new hardwa…
Anthropic Launches AI Research Institute for Societal Impact
Anthropic establishes The Anthropic Institute (TAI) to study AI's real-world effects across economics, security, psychol…