AI Daily Digest
Your daily briefing on AI developments that matter to your business.
🎧 Listen to this digestSunday, July 26, 2026
Major AI Players
5Kimi K3 Open Weights Drop on Hugging Face: The Largest Open-Weight Model Ever at 2.8 Trillion Parameters
Moonshot AI has released the full model weights for Kimi K3 — at 2.8 trillion parameters, the largest open-weight AI model ever published — on Hugging Face. Teams can now self-host the model, keeping prompts off Moonshot's servers and outside the reach of China's National Intelligence Law. The model ranked first on the Frontend Code Arena leaderboard and fourth on the Artificial Analysis Intelligence Index, placing it ahead of Claude Opus 4.8.
Tech TimesLawsuit Filed Against Grok AI Over AI-Generated Child Sexual Abuse Material With Unfiltered 'Spicy' Mode
Potts Law Firm has filed a civil lawsuit in federal court in Little Rock, Arkansas, against the company behind Grok AI, alleging its 'spicy' and 'unhinged' modes — which remove safety filters — enabled a photographer to generate AI-produced child sexual abuse material depicting real local children. The firm says more lawsuits are planned for the following week on behalf of additional affected families. Grok AI did self-report the perpetrator to authorities in June before the civil suit was filed.
KHBS / 40/29 TVAI Industry Has Spent $65M on U.S. Midterm Elections, With Anthropic, OpenAI, and Meta-Linked Super PACs Leading the Push
The AI industry has already committed at least $65 million to influence the 2026 U.S. midterm elections through a network of super PACs, with Anthropic, Meta, and OpenAI President Greg Brockman among those linked to outside spending groups. The largest vehicle is 'Leading the Future,' a pro-AI super PAC network associated with Silicon Valley elites that has redirected $20 million to a Republican-focused PAC. The spending reflects the industry's effort to shape the next Congress on issues including AI regulation, open-weight model restrictions, and export controls.
Business InsiderSpaceX Acquires AI Coding Tool Cursor for $60 Billion in All-Stock Deal
SpaceX has acquired Cursor, the AI-native code editor built by Anysphere and founded by four MIT students, in a $60 billion all-stock deal shortly after SpaceX's IPO. Cursor, which is forked from Visual Studio Code and offers AI chat, agentic coding, and codebase-awareness features, gives SpaceX and its Grok Build initiative direct distribution to millions of professional developers. The acquisition connects Cursor's developer user base to SpaceX's Colossus supercomputer for AI model training.
ForbesAI Distillation Moves From Technical Footnote to Washington Policy Crisis as China Rivalry Intensifies
The technique of AI distillation — training a smaller model using outputs from a larger frontier model — has rapidly moved from an obscure machine learning method to the center of a Washington policy battle over China's AI rise. Chinese labs like Moonshot are releasing powerful open-weight models partly enabled by distillation from U.S. frontier models, which the White House has characterized as potential theft. U.S. policymakers are now debating whether to restrict distillation as part of broader AI export controls, while major tech companies argue such restrictions would harm open-source innovation.
CNBCGlobal Developments
2Japan Launches National Physical AI Infrastructure Program, Targets 30% of Global AI Robotics Market by 2040
Japan's Ministry of Economy, Trade and Industry (METI) has launched the FRONTia Project — described as the world's first national AI infrastructure dedicated to physical AI — to develop multimodal foundation models for robotics and industrial automation. Announced at Nvidia's AI conference in Japan, the initiative brings together domestic manufacturers and robotics firms to build open-world AI models for industrial applications. Japan has set an ambitious target of capturing more than 30% of the global AI robotics market by 2040.
The Manila TimesChinese AI Models Are Winning U.S. Users on Price and Performance as Kimi K3 and Z.ai GLM-5 Go Mainstream
A growing number of American professionals, including Mozilla's CTO, have switched to Chinese AI models like Moonshot's Kimi K3 for daily tasks, citing near-frontier performance at a fraction of the cost of OpenAI or Anthropic. The trend reflects a maturing China-U.S. AI race where Chinese labs are now competitive on capability benchmarks, not just price. Broader adoption is complicated by data sovereignty concerns and U.S. government scrutiny over distillation and model theft allegations.
AP NewsRobotics
4Mitsubishi Motors Partners With University of Tokyo Startup Highlanders to Mass-Produce Humanoid Robots at 1,000 Units Per Month by End of 2027
Mitsubishi Motors has teamed up with Highlanders Inc., a University of Tokyo spinout, to develop and mass-produce AI-powered humanoid robots, with a target of 1,000 units per month by end of 2027. The automaker plans to deploy the robots on its own engine manufacturing line in Kyoto to collect production data. Mitsubishi's CEO positioned the company as Japan's first mover in humanoid mass production, directly competing with global players like Figure, Agility, and Tesla.
Japan Wire by Kyodo NewsJohnson & Johnson Reveals Ottava Surgical Robot With Space-Saving Under-Table Arm Design
Johnson & Johnson has publicly unveiled its Ottava surgical robot, which features four robotic arms that extend from beneath the operating table rather than from a ceiling boom or cart — a design J&J says occupies 30–50% less operating room space than traditional systems. The company says it has redesigned the most commonly used surgical instruments from the ground up, including a new two-in-one needle driver. The system positions J&J as a direct competitor to Intuitive Surgical's da Vinci platform.
The Robot ReportIEEE Humanoids 2026 Conference Names 'Future of Work' as Defining Theme as Commercial Deployments Accelerate
The IEEE-RAS International Conference on Humanoid Robots, set for December 6–9 in Santa Clara, has officially designated 'Humanoids and the Future of Work' as its 25th anniversary special topic — a signal of how quickly the field has shifted from research to commercial reality. Agility Robotics' Digit is already working commercial shifts at Toyota's Ontario factory and a GXO Logistics warehouse, following Agility's $2.5 billion SPAC merger in June 2026. The conference will serve as a key gathering point for an industry now wrestling with labor displacement questions at scale.
Tech TimesHumanoid Robot Pilot Draws Teacher Backlash at New York High School on Seneca Nation Territory
Salamanca High School in New York is set to become the first U.S. school to deploy a humanoid robot and AI teaching assistant, supplied by Realbotix — a company that previously acquired a sex doll manufacturer. Teachers and the New York State United Teachers union have raised serious concerns about student data privacy, software security, and the appropriateness of the vendor choice given the school's large Native American student population. The rollout is proceeding as part of a district pilot program despite the opposition.
The GuardianHealthcare & Biotech
1HKU Researchers Develop ClairS Deep-Learning Algorithm That Improves Cancer Mutation Detection Across Multiple Cancer Types
A research team at the University of Hong Kong has developed ClairS, a deep-learning algorithm that significantly improves detection of somatic cancer mutations using long-read DNA sequencing. The team addressed the chronic shortage of real cancer training data by generating synthetic tumor-normal data from normal human samples, enabling high-accuracy results across breast cancer, lung cancer, and melanoma datasets. The researchers say the approach offers a scalable blueprint for training medical AI when real clinical data is scarce.
University of Hong KongInvestment Noteworthy
3Anthropic Files for October IPO at $965B Valuation, Tapping Morgan Stanley, Goldman, and JPMorgan
Anthropic has mandated Morgan Stanley, Goldman Sachs, and JPMorgan to underwrite an IPO targeting a $965 billion valuation, with a listing expected in October 2026. The filing comes the same week the company launched Claude Opus 5, its most capable model to date. Analysts note Anthropic is positioning itself not just as an AI model provider but as safety-first AI infrastructure with utility-like characteristics — a framing designed to justify its premium multiple.
FourWeekMBAElon Musk's Boring Company in Talks to Raise $4 Billion at $20 Billion Valuation
The Boring Company is in discussions to raise a $4 billion funding round at a $20 billion valuation, more than triple its 2022 valuation of $5.7 billion, according to The Wall Street Journal. The startup has already built tunnel networks under Las Vegas and has announced projects in Nashville and Dubai, with additional pitches underway in Baltimore, Chicago, and Los Angeles. The round reflects renewed investor appetite for Musk-adjacent infrastructure ventures following SpaceX's recent IPO.
TechCrunchDeepSeek Pauses Its Second Fundraising Round After Founder's Comments About U.S.-China AI Competition Go Viral
DeepSeek has verbally informed prospective investors that it is suspending its second fundraising round, which had been targeting at least 10 billion yuan (~$1.4B), after comments attributed to founder Liang Wenfeng about U.S.-China AI competition leaked online and went viral. The suspension stems partly from Liang's frustration over reports from his first financing deal, which closed in June and raised $7 billion. DeepSeek has also begun preparations for an IPO that could file as soon as later this year.
BloombergGeneral AI
2Coca-Cola Deploys Project Fizzion, an Adobe-Built AI System, to Govern Brand Consistency Across 200 Markets
Coca-Cola has launched Project Fizzion, an AI system co-developed with Adobe that converts the company's brand guidelines into a machine-readable 'StyleID' format, allowing designers in any of its 200+ markets to create localized campaigns that automatically adhere to global brand standards in real time. The system replaces an extensive manual review process that previously applied to the 2.2 billion daily servings Coca-Cola's brands touch worldwide. It represents one of the first large-scale deployments of AI as an always-on brand governance layer rather than a creative generation tool.
ForbesNew Research Finds AI Mental Health Advice Is Routinely Misjudged Because Evaluations Ignore Conversational Context
A new study highlights a fundamental flaw in how AI mental health tools are evaluated: most assessments test models in a 'stateless' way — single prompts with no prior conversation — while real users engage in ongoing, contextual dialogues where advice quality changes dramatically. The research found that AI responses rated acceptable in isolation can become harmful or inappropriate when evaluated in the context of a full conversation. The findings have significant implications for any business deploying AI in sensitive advisory roles.
Forbes