Anthropic's Claude Sonnet 4.5 achieves 94% political neutrality rating, outperforming competitors GPT-5 and Llama 4, as the company faces Trump administration accusations of liberal bias and rewrites system instructions to ensure balanced political responses.
Sakana AI shatters Japanese startup records by achieving a massive $2.6 billion valuation, officially becoming Japan's most valuable unicorn company and cementing its position as the nation's largest unlisted technology startup following its latest funding round worth approximately 400 billion yen.
ByteDance launches Depth Anything 3, a groundbreaking AI model that transforms any visual input into accurate 3D geometry using a single transformer backbone, dramatically outperforming previous models across depth estimation and 3D generation tasks while offering open-source access with multiple variants and interactive tools.
AI experts remain divided on whether large language models demonstrate genuine intelligence or merely sophisticated pattern recognition, as these systems successfully tackle complex problems like Math Olympiad questions while working with word-embedded knowledge, though multimodal AI models face greater skepticism due to their pixel-based processing methods.
Researchers discover that AI models dramatically switch between accurate and false responses based solely on how questions are phrased, with 'think step by step' prompts triggering correct factual recall while 'give answer in one word' instructions activate shallow processing circuits that frequently generate hallucinated information.
Baidu unveils ERNIE 5.0 multimodal AI model claiming to outperform GPT-5 and Gemini 2.5 Pro in visual tasks while pricing competitively at $0.85-$3.40 per million tokens as the Chinese tech giant expands globally with AI tools that already power 83% of livestreamers during major shopping events.
Google DeepMind unveils SIMA 2, a groundbreaking AI gaming companion powered by Gemini that evolves from following simple instructions to reasoning, conversing, and self-improving through trial-and-error in 3D virtual worlds, demonstrating unprecedented adaptability across unseen games while understanding sketches, emojis, and multiple languages.
New research testing 24 advanced AI language models reveals they systematically fail to distinguish between belief and knowledge, with accuracy plummeting from over 90% to as low as 14.4% when processing first-person false beliefs, exposing critical flaws in their reasoning capabilities.
OpenAI releases GPT-5.1 with two variants - GPT-5.1 Instant for enhanced conversations and GPT-5.1 Thinking for adaptive reasoning - while introducing specialized mental health safety features targeting delusions, psychosis, and emotional dependency concerns.
World Labs launches Marble AI platform that transforms text prompts, photos, and videos into persistent, downloadable 3D environments for gaming and VR industries, offering subscription plans from free to $95/month with unique AI-native editing tools that separate spatial structure from visual style.