Safety
AI and Data Centers Outpace Israel, Racism, and Crypto in US Campaigns
AI dominates US campaign discourse, appearing in 40% of House, Senate, and governor races, per Washington Post analysis.
LATEST NEWS
-
Intel's New CPU Outperforms NVIDIA in AI Workloads
Intel's latest CPU achieves 1.6x speed boost over NVIDIA's GPU in AI tasks, signaling a shift in hardware preferences.
-
Epoch AI Survey Shows 20% of US Workers Delegate Tasks to AI
A survey by Epoch AI and Ipsos found that 20% of employed Americans now delegate at least one task to AI instead of colleagues, according to data from July 2026.
-
Artificial Analysis Launches Optima to Address AI Benchmarking Limitations
Artificial Analysis has released Optima, a platform enabling users to build custom benchmarks for AI models, with early tests showing cost reductions of up to 10 times in finance and accounting applications.
-
Anthropic's Bio-Weapons Filter Down for 11 Months
Anthropic's biological weapons filters were offline for nearly a year, exposing 133 million requests during the period.
-
OpenAI Dissolves Team to Monitor Catastrophic AI Risks
OpenAI disbanded its Preparedness team in July 2026, shifting its focus to other groups, according to Financial Times.
-
Google Study Reveals AI Models' Beliefs Shift When Denied Self-Reflection
A Google-led study found that disabling AI models' ability to deny consciousness alters their views on animals, religion, and human-like traits, with scores rising to 7.5 for animal sentience.
-
Mathematicians Say LLMs Excel at Calculations but Struggle with Creativity
Top mathematicians argue large language models are strong at solving known problems but lack the ability to generate novel ideas, according to a recent analysis.
-
Woman Alleges Stepfather Used Grok to Create Explicit Images
A woman claims her stepfather used Grok to generate over 7,000 explicit images of her from a childhood photo, leading to a lawsuit against xAI.
-
SpaceX Completes Acquisition of AI Coding Startup Cursor
SpaceX has finalized its $60 billion acquisition of Cursor, an AI coding startup, following a deal announced in April.
-
Anthropic Details How Watermarks Will Work on Claude Output
Anthropic outlined how its new watermarking system will identify AI-generated text, noting that light editing won’t fully remove the watermark.
-
txtai Introduces LEMUR for Late-Interaction Retrieval
txtai's LEMUR model improves retrieval performance by 56.6% on nfcorpus, according to benchmark results from August 2026.