Neura News

AI News

News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.

Latest News

25 articles

Research

Modular Pretraining: A New Approach to Containing Dangerous AI Knowledge

Researchers at Anthropic and AE Studio have introduced Gradient Routed Auxiliary Modules (GRAM), a method that isolates dangerous knowledge in large language models into switchable modules during training. This approach allows operators to control access to sensitive content, potentially reducing risks of misuse. Preliminary experiments show promise across models up to 5B parameters, but the method has not yet been applied to production-scale systems.

Aug 1712 minNeura News

Industry

Amazon Turns Twitch Streams Into Training Data for Generative AI

Amazon is using its live-streaming platform Twitch to train generative AI models, according to a brief BBC report. The report, published as a video description, notes that Twitch's audio, video, and chat data could serve as a rich dataset for AI development. No further details or company comments were provided.

Aug 135 minNeura News

Developer

h3-metal Brings MiniMax-H3 Video Generation to Apple Silicon

h3-metal is a new open-source project that brings MiniMax-H3 text-to-video generation to Apple Silicon via native Metal. It supports prompt-to-video, prompt-to-audio, first/last frame conditioning, and Ref2VA references. The release includes performance optimizations, quality tradeoffs, and a tutorial for building fast videos on M3 Max and M5 Max hardware.

Aug 1123 minNeura News

AI Models

Google DeepMind Unveils Gemini Robotics 2 and ER 2 for Smarter Machines

Google DeepMind has unveiled Gemini Robotics 2, a vision-language-action model for robot control, and Gemini Robotics ER 2, an embodied reasoning model. The former handles direct motor commands, while the latter plans actions. ER 2 is available in Google AI Studio, while Robotics 2 is on a waitlist.

Jul 314 minNeura News

Funding

Fish Audio Raises $50M Seed for AI Voice Models

Fish Audio, a Palo Alto-based startup building AI voice models, has raised $50 million in a seed round led by Coreline Ventures and Capital Today. The company offers over 15,000 natural language controls and serves more than 8 million users, generating $21 million in annual recurring revenue. It plans to use the funding to develop more advanced models and expand enterprise offerings.

Jul 285 minNeura News

AI Models

Flux 3 generates 20-second videos with native audio

German AI company Black Forest Labs released Flux 3, a multimodal foundation model trained on images, video, and audio together. It can generate videos up to 20 seconds long with native audio for the first time. In early tests, it beat several rivals including Luma Ray 3.2 and Runway Gen-4.5. The company also developed Flux-mimic, a video-action model for robotics, now being tested at Audi.

Jul 233 minNeura News

Funding

Microsoft and Mistral in Multi-Billion-Dollar AI Infrastructure Deal

Microsoft and Mistral have expanded their strategic partnership with a multi-billion-dollar agreement to build AI infrastructure across Europe. Mistral will deploy thousands of Nvidia Vera Rubin GPUs, and Microsoft will use that capacity for its cloud and AI services. Mistral's Medium 3.5 and OCR 4 models are now available in Microsoft Foundry, with Medium 3.5 also in Copilot Studio. The deal targets regulated industries like finance, healthcare, and manufacturing, allowing companies to run Mistral models through Azure Local in the cloud, on-premises, or offline. Microsoft President Brad Smith emphasized Europe's need for powerful AI without sacrificing data control.

Jul 213 minNeura News

AI Models

Alibaba Qwen Audio 3.0 TTS Plus Tops Speech Arena Leaderboard

Alibaba's Qwen-Audio-3.0-TTS-Plus has claimed the top spot on Artificial Analysis' Speech Arena leaderboard for provider voices, achieving an Elo score of 1,236. The model narrowly beats SpeechifyAI's Simba 3.2 (1,234) and other competitors like Gemini 3.1 Flash TTS and Sonic 3.5. Available in Flash and Plus versions, it supports 16 languages and offers natural language style control, though its speed lags behind rivals.

Jul 212 minNeura News

Industry

Sony Sues Udio Over 30,000 Songs in AI Copyright Case

Sony Music Entertainment has filed a new lawsuit against AI music generator Udio, accusing it of infringing copyright on more than 30,000 songs. The list includes tracks from Beyoncé, Britney Spears, Johnny Cash, and Harry Styles. Sony claims the songs were copied and ingested into Udio's generative AI models without permission.

Jul 203 minNeura News

AI Models

Thinking Machines Lab Releases Open-Weight Inkling AI Model

Mira Murati's Thinking Machines Lab has released Inkling, its first open-weights AI model. The 975B parameter Mixture-of-Experts transformer is Apache-2.0 licensed and trained on 45 trillion tokens of text, images, audio, and video. While not a frontier model, it aims to serve as a strong base for fine-tuning via the Tinker platform.

Jul 163 minNeura News

AI Models

AI Voice Fraud Outpaces Defenses as Losses Hit $893 Million

The FBI reported over $893 million in AI-enabled fraud losses in 2025, with older adults losing $352 million. Voice cloning requires just three seconds of audio, and detection experts admit they can no longer reliably distinguish real from fake. Banks, telecoms, and regulators are struggling to contain a threat that is both technically advanced and emotionally devastating.

Jul 1525 minNeura News

AI Tools

One-Click Integration from Hugging Face to SageMaker Studio

Hugging Face and Amazon SageMaker AI have launched a deep-link integration that lets developers go from model discovery on Hugging Face directly into SageMaker Studio with a single click. The integration includes pre-configured permissions and GPU quota visibility, reducing the friction of manual setup for fine-tuning and deployment.

Jul 74 minNeura News