AI News Roundup – Trump administration plans new AI guidelines for federal agencies, AI use found to add to workers’ burdens, Nvidia planning to develop open-source AI models, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • The Trump administration is drafting new guidelines regarding AI contracts for federal agencies amid its dispute with Anthropic, according to the Financial Times. New draft guidance from the General Services Administration (GSA), which oversees procurement for civilian agencies in the U.S. federal government, would mandate that AI vendors grant the government an irrevocable license to use their AI systems for “any lawful” use. The GSA would also require vendors to provide AI tools that do not “manipulate responses in favor of ideological dogmas such as diversity, equity, inclusion,” in line with administration opposition to so-called “woke” AI. The proposed guidance follows a public split between the Defense Department and AI company Anthropic over the allowed use of the Claude AI system, which eventually led to a ban on government use of Anthropic products (to take effect later this year) and the designation of the company as a supply chain risk, which Anthropic is now challenging in court. Further changes to the GSA guidelines are expected in coming weeks before they are finalized. 
  • The Wall Street Journal reports that AI use at the workplace may be adding to workloads rather than reducing them. According to data from over 164,000 workers collected by ActivTrak, a productivity-tracking company, the time that workers spent on email and workplace messaging apps more than doubled when using AI in comparison to before they used AI tools. Time spent using HR or accounting software also rose by 94%. AI use was also associated with a 9% decrease in time spent on focused, uninterrupted work. An ActivTrak official postulated that any time saved by using AI is immediately repurposed into other tasks, but appears to come at a cost to worker focus. AI backers have often touted the technology’s boosts to worker efficiency and even suggested that AI could lead to a shorter workweek, but those predictions do not appear to be in line with the data on real-world AI use. A study on similar topics released last month found that while AI use can boost productivity in the short term, it can also lead to “cognitive overload, burnout, poorer decision-making, and declining work quality” in the long run. The findings demonstrate the need for caution when adopting AI at a large scale in the workplace, but many companies appear to be pushing forward with AI adoption regardless. 
  • Nvidia is planning to spend billions to develop its own open-source AI models, according to WIRED. In a recent SEC filing, the company revealed it will spend $26 billion over the next five years to build its own AI models. The models will be open-source, meaning the weights within the model are publicly available. This past week, Nvidia released Nemotron 3 Super, a 128-billion parameter open-source model that claims performance similar to OpenAI’s largest open-source models. The move could transition Nvidia from not only the dominant AI chipmaker but also an AI laboratory in its own right, competing directly with partners such as OpenAI and China’s DeepSeek. Indeed, many Chinese companies have released open-source models, which has made them popular among AI startups around the world, and Nvidia’s latest offerings could increase excitement in that space. An Nvidia executive told WIRED that the models will also help the company improve its chips and datacenters: “We build it to stretch our systems and test not just the compute but also the storage and networking, and to kind of build out our hardware architecture roadmap.” 
  • Bloomberg reports that a recent boom of the AI agent OpenClaw in China has placed strains on the government’s efforts to regulate AI technology. OpenClaw, the AI agent behind the Moltbook AI social network that went viral earlier this year (as reported in this AI Roundup), has taken the Chinese AI world by storm, with major domestic players such as Tencent and Alibaba launching AI tools compatible with OpenClaw and local governments offering subsidies for startups that make use of the technology. The central government in Beijing, however, has struggled to keep up with the rapid advances in the technology as it seeks to place stricter restrictions on certain AI activities. One government-affiliated entity stated this past week that it would test OpenClaw and related technologies later this month and develop standards for use. OpenClaw operates autonomously and requires access to data on a host machine, which could possibly lead to security vulnerabilities, and this has led to some skepticism among both the government and consumers. Until then, however, “raising lobsters” (a nickname for installing OpenClaw based on its mascot) is likely to continue in China for the foreseeable future. 
  • A new gig economy job is helping to train AI models that control robots, according to the Los Angeles Times. Hundreds of California residents have a new side job: strapping cameras onto their bodies to capture footage as they go about their daily chores. The footage is then used to train so-called “physical AI” systems that operate robotic arms and other devices that can interact with the physical world. One startup, Scale AI, has gathered over 100,000 hours of such footage, and the market for this sort of data is expected to grow rapidly in the coming years. Some have criticized this manner of collecting data as exploitative, but many workers are turning to these sorts of jobs due to ongoing economic uncertainty. Others are jumping at the chance to capitalize on a newer market: one founder of a data-capture startup said that physical data is “one of the biggest gig economies that is going to exist in the whole world.”