AI News Roundup — August 11, 2026
Open-weight models took center stage as Nvidia revealed a trillion-parameter Nemotron 4 family and Meta pushed its open-weight strategy further, while Google's Gemini app crossed one billion monthly users, Anthropic moved to watermark AI-generated text under new EU rules, and OpenAI lost another senior executive.
Nvidia developing a trillion-parameter open model family, Nemotron 4
Nvidia is reportedly building Nemotron 4, a new open-model family meant to rival leading open-weight systems from both U.S. and Chinese labs. Reuters reports the largest variant could exceed 1 trillion parameters, though training is unfinished and no release date is set, with a possible arrival as early as late fall. Nvidia also shipped Nemotron 3.5 Lightning for code review, tool use, and security monitoring, plus NeMo Switchyard, an open-source model-routing library. "Every company and every country needs accessible frontier open models to strengthen safety and security, accelerate innovation," said Nvidia VP Kari Briski.
Google says the Gemini app has passed one billion monthly users
Google announced the standalone Gemini app has surpassed 1 billion monthly active users, becoming the 14th Google product to hit that scale (the figure excludes Gemini usage inside Search, Workspace, and Android). Google says 63% of users engage via voice, the app generates more than 150 million images daily, and it has over 100 million active iOS users. The app had already topped 950 million users in late July. The milestone puts Gemini roughly alongside ChatGPT, which Reuters reported also crossed 1 billion monthly users in June.
Anthropic to watermark Claude-generated text under new EU rules
Anthropic said it will watermark text generated by Claude models released after August 2 — the date the EU AI Act's transparency code took effect — and plans to extend the system to older models. The watermark will apply across Claude's consumer products, the API, Claude Code, Claude Cowork, and Claude Tag, and is embedded in the text itself so it can travel through copying and pasting, though Anthropic acknowledges its durability after heavy editing is unclear. Files will use the C2PA provenance standard. "Because the watermark is part of the text, it will travel with the text when it's copied and pasted elsewhere, and may persist through some editing," Anthropic's support documentation states.
Manus returns to independent operation as Meta acquisition unwinds
Chinese AI-agent startup Manus said it will resume operating independently after its planned acquisition by Meta was unwound. Reuters reports Beijing ordered Meta to reverse the more than $2 billion deal amid heightened scrutiny of foreign investment in Chinese frontier-technology companies; Tencent has reportedly been discussing becoming Manus's largest shareholder. Manus said some user data generated on or after December 29, 2025 will be deleted later in August to meet regulatory requirements in certain jurisdictions, with affected users getting notice and a chance to back up data. "This is part of our separation from Meta," the company said.
River AI, founded by an xAI co-founder, raises $1.1 billion out of stealth
River AI, founded by former xAI co-founder Igor Babuschkin, raised $1.1 billion in a seed/Series A round led by General Catalyst and AMP PBC, with Nvidia, AMD Ventures, Y Combinator, and Temasek also participating. The company wants to rebuild the AI stack around "personally trainable agents" — open models individuals and enterprises can adapt via reinforcement learning and LoRA fine-tuning — with an eventual push into dedicated hardware for local, personal AI. "The stack has to be rebuilt end to end: training, models, the product layer, and new hardware that lets personal AI live close to you," Babuschkin said.
OpenAI executive Brad Lightcap reported to be departing for a new venture
Reuters reports OpenAI executive Brad Lightcap, a senior operational and strategic figure at the company, is leaving to pursue a new venture. The departure adds to a string of high-profile exits from OpenAI's leadership ranks this year and comes as the company navigates litigation, a planned IPO process, and intensifying competition.
Meta pushes its open-weight strategy with Muse Glimmer; xAI launches always-on Grok Bot
Meta launched an open-weight model called Muse Glimmer and renewed its broader open-weight push, with Mark Zuckerberg arguing for sharing AI model checkpoints more widely, including potentially with the U.S. government. Separately, xAI introduced Grok Bot, an always-on agent for macOS and iOS designed to operate applications and carry out persistent tasks on a user's behalf.