The news desk
Recent developments in artificial intelligence, with the original sources one click away.
President Trump has directed all US federal agencies to cease using Anthropic's AI technology following a clash with the Pentagon over AI safety and military applications.
Sources · Trump directs all federal agencies to stop using AI company Anthropic's technology / Trump orders US agencies to stop use of Anthropic technology amid dispute over ethics of AI - The Guardian +15
About 12% of American teenagers are turning to general-purpose AI tools for emotional support, concerning mental health professionals.
Sources · TechCrunch AI
Amazon demonstrates efficient serving of multiple fine-tuned models using vLLM with LoRA adapters on SageMaker and Bedrock platforms.
Sources · Amazon AWS AI
Software engineer Riley Walz is joining OpenAI to develop new human-AI interaction methods, according to WIRED.
Sources · WIRED AI
Researchers publish three arXiv papers tackling large language model optimization via skill fusion, energy-based interpretation, and rank-sparsity compression.
Sources · arXiv AI (cs.AI)
Recent arXiv preprints examine autonomous geoscience data discovery, security risks in multi-agent coordination, and economic theory for AI systems.
Sources · arXiv AI (cs.AI)
AWS announces global cross-region inference for Claude models now available in seven new countries across Asia-Pacific and Middle East regions.
Sources · Amazon AWS AI
Anthropic releases new Claude Sonnet 4.6 model while facing Pentagon tensions over AI safeguards, as Google launches Gemini 3.1 Pro.
Sources · Last Week in AI
AI firms in India are focusing on building large user bases through free offerings as they test monetization potential in the market.
Sources · TechCrunch AI
Author Michael Pollan contends in 'A World Appears' that AI lacks the capacity for consciousness despite its capabilities.
Sources · WIRED AI
Perplexity AI launches 'Perplexity Computer,' a new unified platform designed to manage projects from research to deployment using multiple AI models.
Sources · Perplexity AI Launches Computer Platform for Unified AI-Driven Project Management / Perplexity launches Computer to streamline end-to-end AI projects - Crypto Briefing +7
New research identifies challenges in vision-language models including long-horizon tasks, causal reasoning deficiencies, and artifact propagation.
Sources · arXiv AI (cs.AI)
Anthropic claims DeepSeek, Moonshot, and MiniMax used 24,000 fake accounts to extract Claude's AI capabilities for training their own chatbots.
Sources · NYT Technology / TechCrunch AI
Meta has announced a colossal AI chip deal with AMD worth over $100 billion, including a stock incentive, alongside its ongoing Nvidia partnership and in-house chip development to power its ambitious AI goals.
Sources · Tom's Hardware / Smart Forum +12
Companies are hiding the extent of human involvement in operating supposedly autonomous humanoid robots, according to MIT Technology Review.
Sources · MIT Technology Review
Hugging Face releases smolagents library for building AI agents, while researchers develop frameworks for ranking systems and fish identification.
Sources · Amazon AWS AI / arXiv AI (cs.AI)
Three arXiv papers investigate LLM layer effectiveness, energy distribution during inference, and unequal language comprehension abilities.
Sources · arXiv AI (cs.AI)
Researchers publish papers on improving LLM judges with tools, user-controlled recommendations, and dynamic resource allocation in RL.
Sources · arXiv AI (cs.AI)
India's AI Impact Summit brings together leaders from OpenAI, Anthropic, Nvidia, Microsoft, Google, Cloudflare, and government officials this week.
Sources · TechCrunch AI
Recent arXiv preprints explore scaling code language models, improving LLM reasoning through confidence-driven decoding, and question answering over long audio.
Sources · arXiv AI (cs.AI)