Justice Department Defends Penalty Against Anthropic Over Military AI Use Restrictions
DOJ says it lawfully penalized Anthropic for attempting to limit military use of its Claude AI models, according to court filing.
Sources · WIRED AI
The news desk
Recent developments in artificial intelligence, with the original sources one click away.
DOJ says it lawfully penalized Anthropic for attempting to limit military use of its Claude AI models, according to court filing.
Sources · WIRED AI
New neuro-symbolic architecture enables multimodal language models to reason about complex physical systems without domain-specific training.
Sources · arxiv.org
OpenAI introduces smaller, faster GPT-5.4 variants optimized for coding, tool use, and high-volume API workloads.
Sources · OpenAI
The Pentagon is developing plans to allow AI companies to train models on classified military data in secure environments.
Sources · MIT Technology Review
Research shows GPT-4.1 exhibits risk behaviors predicted by Prospect Theory when assigned different socioeconomic personas in gambling scenarios.
Sources · arxiv.org
Amazon Web Services announces disaggregated inference capabilities on SageMaker HyperPod EKS, enabling separated compute and memory resources for AI models.
Sources · Amazon AWS AI
A Florida AI regulation bill backed by Gov. DeSantis failed after President Trump signaled opposition to state-level tech restrictions.
Sources · NYT Technology
Research finds combining different AI models outperforms single-vendor teams in medical diagnosis by pooling complementary strengths.
Sources · arxiv.org
MIT Technology Review reports on concerns about OpenAI's technology reaching Iran following its Pentagon agreement.
Sources · MIT Technology Review
New research examines efficient coordination of AI agents and reveals safety vulnerabilities in test-time learning methods.
Sources · arxiv.org
A lawyer handling AI-related cases says chatbots are now connected to mass casualty incidents, with technology outpacing safety measures.
Sources · TechCrunch AI
Recent studies expose fragility in LLM reasoning agents and introduce frameworks for chemistry, networks, and streaming video applications.
Sources · arxiv.org
Study introduces OffTopicEval benchmark, finding even top-performing LLMs fall short on operational safety with scores below 80%.
Sources · arxiv.org
NVIDIA unveils hybrid AI model, XAI loses another cofounder, and Anthropic sues the Department of Defense in recent AI developments.
Sources · Last Week in AI
New research introduces SPARROW, a model that improves spatial precision and temporal consistency in video analysis through tracked features.
Sources · arxiv.org
ChatGPT now allows users to access Spotify, Canva, Figma, Expedia, DoorDash, and Uber directly within the chat interface.
Sources · TechCrunch AI
AWS has integrated P-EAGLE, a parallel speculative decoding technique, into vLLM v0.16.0 to accelerate large language model inference.
Sources · Amazon AWS AI
Google's generative AI search tools are citing the company's own services more frequently than third-party publishers, according to WIRED.
Sources · WIRED AI
A lawyer handling AI-related cases reports chatbots are now linked to mass casualty incidents, with technology advancing faster than safety measures.
Sources · TechCrunch AI
Father-son founded startup Nyne secures seed funding to build data infrastructure that gives AI agents missing human context.
Sources · TechCrunch AI