Google Taps Samsung for Next-Gen AI Chip as TSMC Capacity Tightens

Google DeepMind and partners have announced a $10M funding call dedicated to multi-agent AI safety research. The initiative aims to address safety challenges that emerge when multiple AI systems interact with each other. This represents a targeted investment in a research area that has received less attention than single-agent safety concerns.

A German court ruled that Google is legally responsible for the accuracy of content generated by its AI Overviews feature, which produces AI-generated answers within Google search results. The ruling treats AI-generated content as Google's own statements rather than neutral search results, establishing potential liability for factual errors. This decision could have broad implications for how AI-generated content is regulated across jurisdictions.

Google DeepMind released DiffusionGemma, a 26B Mixture of Experts model that generates text up to 4x faster than autoregressive models by producing entire blocks of text simultaneously rather than token-by-token. The open experimental model, available under Apache 2.0 license, achieves 1000+ tokens per second on NVIDIA H100 GPUs and fits within 18GB VRAM on consumer hardware when quantized. The trade-off is lower output quality compared to standard Gemma 4, positioning it for speed-critical applications like real-time editing and code infilling rather than production use cases demanding maximum quality.

Google DeepMind introduced Gemma 4 12B, a multimodal AI model designed to run on consumer laptops with 16GB of RAM. The model uses an encoder-free architecture that processes vision and audio inputs directly into the language model backbone, reducing latency and memory overhead. Performance approaches the larger 26B model while maintaining a smaller footprint, and it is released under an Apache 2.0 license.