AI & Machine Learning

The French AI startup ships its first text-to-speech model under open weights, offering enterprises voice cloning from three seconds of audio and 70-millisecond latency across nine languages.

Google Research unveils TurboQuant, a training-free compression algorithm that reduces LLM key-value cache memory by 6x and boosts inference throughput up to 8x, sending SK Hynix and Samsung shares down 5-6 percent.

OpenAI discontinues its Sora video generation app, ends a three-year Disney licensing deal, and redirects compute to a new model codenamed Spud ahead of a potential IPO.

The U.S. Department of Energy announced a partnership with SoftBank subsidiary SB Energy to develop a 10 GW data center and matching power generation complex at the decommissioned Portsmouth uranium enrichment plant in Pike County, Ohio, backed by $33.3 billion in Japanese investment.

Google added Cart, Catalog, and Identity Linking capabilities to its Universal Commerce Protocol, expanding the open standard that lets AI agents complete purchases across retailer platforms.

IBM Research, Red Hat, and Google Cloud donated llm-d, a Kubernetes-native LLM inference framework, to the CNCF as a sandbox project at KubeCon Europe 2026, backed by NVIDIA, CoreWeave, AMD, and other partners.

Anthropic has launched a research preview enabling Claude to operate Mac desktops directly, opening files, navigating browsers, and executing tasks across applications with a permission-based safety model, available to Pro and Max subscribers.

An anonymous trillion-parameter AI model that topped OpenRouter usage charts and triggered widespread speculation about a stealth DeepSeek V4 launch turned out to be Xiaomi's MiMo-V2-Pro, signaling the smartphone and EV maker's escalating ambitions in the AI model race.

MIT researchers present Wave-Former and RISE, two systems that combine millimeter-wave radar with generative AI to reconstruct hidden objects and entire indoor scenes through obstructions, achieving up to twice the precision of prior methods.

NVIDIA, Siemens, and Synopsys each launched digital twin platforms in March 2026, targeting AI factory simulation, adaptive manufacturing, and automotive electronics virtualization.

Sony Music's disclosure that it has flagged more than 135,000 AI-generated deepfake tracks for removal exposes a widening split among streaming platforms over how to detect, label, and restrict synthetic music, with approaches ranging from voluntary self-reporting to automated detection.

Mistral releases Small 4, a 119B-parameter open-weight model under Apache 2.0, and Forge, an enterprise platform for training custom models on proprietary data, as open-weight AI closes to within three months of frontier systems.