AI Daily Digest
Your daily briefing on AI developments that matter to your business.
🎧 Listen to this digestMonday, July 13, 2026
Government & Policy
3FDA Moves to Allow AI Models in Clinical Trials and Single Late-Stage Trial Approvals
The FDA has unveiled a broad initiative to accelerate drug approval, including allowing AI models to be used in the clinical trial process and permitting single late-stage trials (instead of the traditional two) in certain cases. The reforms, announced last month under HHS Secretary RFK Jr., are designed to maintain U.S. leadership in drug development and speed patient access to new therapies. The changes represent a significant regulatory opening for AI-driven drug discovery companies.
Los Angeles Business Journal69% of Americans Support Bernie Sanders Plan to Tax AI Companies 50% in Stock for Public Sovereign Wealth Fund
A new Verasight poll finds 69% of Americans support Senator Bernie Sanders' proposal to require major AI companies to transfer 50% of their stock to a public sovereign wealth fund — giving citizens direct ownership stakes in the AI industry. Sanders has framed the plan as a response to AI companies extracting value from publicly funded research and the broader internet. The bill would apply to companies like OpenAI, Anthropic, and xAI, several of which are preparing IPOs.
GizmodoSouth Korea Announces Record $531B Budget Fueled by AI Chip Tax Revenue
South Korea announced it will draft a record budget of over 800 trillion won ($531 billion) for fiscal 2027, with the surge in tax revenues driven largely by the booming AI chip industry. The government plans to restructure 50 trillion won in spending — double last year's level — to fund strategic priorities. The announcement underscores how AI hardware demand is now reshaping national fiscal policy in chip-producing nations.
ReutersMajor AI Players
8OpenAI's GPT-5.6, xAI's Grok 4.5, and ChatGPT Work Agent Launch in Wave of Frontier AI Releases
American AI labs flooded the zone this week with major releases: OpenAI's GPT-5.6 family (including budget model Luna at one-fifth the predecessor's price), xAI's Grok 4.5, and the fully agentic ChatGPT Work interface that can 'take action across your apps and files, stay with a project for hours.' OpenAI CEO Sam Altman declared GPT-5.6 Sol 'the best model in the world right now,' while the new ChatGPT Work agent is designed to turn a goal into finished work autonomously. OpenAI is also launching a dedicated keyboard next week featuring a single large microphone button.
AxiosTencent's Hy3 Model Bets on Agentic Efficiency Over Raw Scale
Tencent launched Hy3, a 295-billion-parameter Mixture-of-Experts model with only 21 billion active parameters, optimized for real-world AI agents and enterprise productivity rather than benchmark supremacy. Independent evaluations show Hy3 is competitive with Claude Opus 4.8 and GPT-5.5 on agentic search and tool orchestration, with a low 5.4% hallucination rate. The release reflects a broader Chinese AI strategy of prioritizing deployment efficiency over model size due to hardware constraints.
ForbesChina's GLM-5.2 Open-Source Model Draws DeepSeek Comparisons With 1M-Token Context Window
GLM-5.2, an open-source model from Beijing-based Z.ai, is generating significant buzz among developers and investors for its coding and agentic AI capabilities, including a 1 million-token context window comparable to GPT-5.5 and Claude Opus 4.8. Hands-on testing found it capable but slow under load, with wait times of several minutes. The model is free to use, positioning it as another low-cost Chinese alternative to Western frontier models.
Business InsiderCheaper AI Tokens Don't Mean Cheaper Enterprise Agents — Finance Teams Warned to Model Outcomes, Not Prices
A new Forbes analysis warns that falling token prices from Meta, OpenAI, and DeepSeek don't automatically translate to lower enterprise AI costs, because the number of calls per completed task — not the price per million tokens — drives actual bills. OpenAI's new GPT-5.6 family ranges from $1/million tokens for Luna to $30/million for flagship Sol output, while Meta's Muse Spark 1.1 API is priced below both. Finance teams are advised to model cost-per-workflow-outcome rather than advertised rates.
Forbes200+ Economists and Nobel Laureates Call for Urgent Policy Action on AI's Economic Impact
More than 200 researchers and economists — including 15 Nobel laureates and researchers from OpenAI, Anthropic, and Google — have issued a joint statement calling on governments and tech leaders to immediately build institutions and policies to manage AI's economic disruption, particularly large-scale job displacement. The statement calls for deeper research into AI's economic effects and proactive policy frameworks before impacts become irreversible. It represents one of the broadest expert coalitions yet to weigh in on AI's societal consequences.
ReutersSatya Nadella Calls Out AI Labs as Hypocrites on Distillation — China's Exploitation of the Practice Escalates
Microsoft CEO Satya Nadella publicly called it 'ironic' that AI labs claim fair-use rights to train on public web data while simultaneously banning competitors from distilling their model outputs. The comment comes as Anthropic and OpenAI are sounding alarms that China — specifically Alibaba — is using distillation to produce cheap, capable open-source models at a fraction of the cost, with Anthropic briefing Trump national security officials about the threat as recently as this month. Anthropic's 'Mythos' and 'Fable' models were reportedly temporarily taken offline due to White House concerns they could be weaponized.
Business InsiderAnthropic Extends Claude Fable 5 Free Access to July 19 — Twice in One Week
Anthropic has twice extended free access to its flagship Claude Fable 5 model for paid subscribers, first to July 12 and now to July 19, along with a 50% boost to Claude Code weekly rate limits. After July 19, Fable 5 will shift to a credit-based usage model. The repeated last-minute extensions suggest Anthropic is managing compute demand or commercial transition timing carefully.
BleepingComputerAnthropic Hires Monzo Co-Founder Tom Blomfield to Lead Compute Strategy
Tom Blomfield, co-founder of British fintech giant Monzo and a general partner at Y Combinator since 2023, is taking a leave of absence to join Anthropic's compute team. Blomfield cited the criticality of compute access as AI enters 'the early stages of recursive self-improvement.' The hire continues Anthropic's high-profile talent acquisition streak as AI labs compete aggressively for top-tier operators.
Business InsiderRobotics
4STMicroelectronics Takes Stake in Italian Humanoid Developer Oversonic Robotics
Chip giant STMicroelectronics, along with Fondazione ENEA Tech Biomedical and SpotInvest, has acquired stakes in Italian cognitive robotics company Oversonic Robotics, maker of RoBee — billed as the first certified cognitive humanoid robot for complex environments. The investment is aimed at accelerating Oversonic's industrial growth and expanding into the U.S. market. The deal is part of a broader ecosystem-building strategy around cognitive humanoid applications.
Robotics & Automation NewsRobot.com Commercially Launches R-noid Humanoid Under Robot-as-a-Service Model With Nvidia Backing
Robot.com has launched R-noid, a humanoid robot designed for 'repetitive, multi-shift, and hard-to-staff jobs,' under a Robot-as-a-Service (RaaS) model that can go from site visit to autonomous operation in 8–12 weeks. The launch is backed by Nvidia Robotics, Physical Intelligence, and others, and integrates with Robot.com's existing fleet of delivery and transport robots on a unified software stack. The commercial launch signals RaaS is moving from concept to deployable product.
Robotics & Automation NewsHoliday Robotics Raises $103M Series A — Korea's Largest Ever — to Mass-Produce Manufacturing Humanoid
South Korean startup Holiday Robotics closed a $103.4 million (155 billion won) Series A, the largest Series A ever raised by a Korean startup and the biggest single round by a domestic humanoid robot specialist. The company's FRIDAY robot is purpose-built for manufacturing environments, and Holiday Robotics plans to begin commercialization in H2 2026, targeting 100 units/year production in 2027 scaling to 1,000/year. The founding CEO previously built and sold industrial AI vision company Sualab.
Wowtale1X Says Its New NEO Robotic Hand Has 'Solved' the Humanoid Hands Problem
Humanoid robot company 1X unveiled a new robotic hand for its NEO home robot that it claims matches or exceeds human hand performance, with 1X's head of product saying 'the short answer is yes' when asked if the company has solved what Elon Musk called the 'hands problem.' The hand uses sensors, motors, and software to replicate the dexterity of human skin and muscle. Solving dexterous manipulation has been considered one of the core remaining barriers to general-purpose humanoid robots.
Business InsiderTransportation
4Tesla Expands Autonomous Robotaxi Operations to Miami — Its Fifth Market
Tesla has launched commercial autonomous ride-hailing in Miami, its fifth metropolitan market and first expansion outside Texas and California. The service uses Model Y vehicles with Tesla's camera-only Full Self-Driving system, putting it in direct competition with Waymo (which is also operating in Miami) and Zoox. The Miami launch marks Tesla's most aggressive robotaxi rollout yet as the vision-only approach faces scrutiny in dense urban environments.
AUTO Connected Car NewsWaymo Calls Police on Misbehaving Teen Passengers — and Delivers Them to Officers
A Waymo robotaxi in San Mateo, California called emergency services after two 15-year-olds fired gel blaster guns and drank alcohol inside the vehicle, then stopped and held the vehicle until police arrived to detain them. The incident is generating debate about privacy, surveillance, and the extent of autonomous vehicle monitoring inside the cabin. A separate incident in East Hollywood saw a man vandalize a Waymo in the middle of a busy intersection, drawing an LAPD arrest.
News.com.auWired Investigation: Uber's Real AV Strategy Is to Slow Autonomous Vehicle Adoption — Not Enable It
A Wired investigation reveals that Uber has been quietly lobbying against autonomous vehicle regulations in multiple states — including Maryland and New York — while publicly positioning itself as AV-neutral. Uber argues AV industry policy proposals are 'largely unworkable' and have failed to address worker issues, and is advocating for 'hybrid networks' where human drivers remain central. The Uber-Waymo partnership in Phoenix has ended, and the Uber-Waymo relationship is under increasing strain.
WiredApple's Failed Self-Driving Car Program Accidentally Invented the Neural Engine That Powers All Its AI
A new report from Mark Gurman reveals that Apple's scrapped autonomous vehicle project left a lasting legacy: the Neural Engine, the on-device AI processing backbone now central to all Apple Silicon chips. Early in the car project, Apple realized it needed powerful on-device AI and developed the Neural Engine — which then migrated to the M-series Mac chips. Apple is now accelerating development of the M7 Ultra, which will support up to 1.5TB of RAM and serve as the basis for a new server product.
The VergeHealthcare & Biotech
2UK AI Fraud Losses Hit £1.28B as Deepfakes and Synthetic Identities Make Scams Undetectable
UK payment fraud losses reached £1.28 billion ($1.74 billion) in 2025, with UK Finance warning that organised criminal groups are increasingly using AI-powered deepfakes, cloned voices, and synthetic identities to bypass identity checks and impersonate trusted individuals. Authorised push payment fraud rose 19% in 2025. Banks say the sophistication of AI-enabled scams is making them significantly harder for consumers and businesses to detect before money is transferred.
International Business Times UKInsilico Medicine Signs $177M AI Drug Discovery Deal With China Medical System for CNS Indication
China Medical System Holdings has partnered with AI drug discovery company Insilico Medicine in a deal worth up to $177 million in milestones, focused on a 'mass-market' central nervous system indication identified by Insilico's PandaOmics AI platform. The specific CNS condition was not disclosed, but the companies say PandaOmics identified an innovative mechanism of action. The deal marks a return partnership, with CMS having previously collaborated with Insilico.
Fierce BiotechInvestment Noteworthy
3Helsing Raises $1.8B at $18B Valuation in Europe's Largest-Ever Defense Tech Funding Round
German AI defense startup Helsing closed a $1.8 billion Series E round at an $18 billion valuation, making it the largest funding round ever for a European defense technology company. Investor demand significantly exceeded available allocation. The round will accelerate Helsing's development of AI platforms for defense applications across partner nations, positioning it as Europe's primary answer to U.S. rival Anduril, which raised $5 billion at a $61 billion valuation in May.
CNBCTSMC Reports Record Q2 Revenue, Up 36% Year-Over-Year on AI Demand
TSMC, the world's largest contract chipmaker, reported record second-quarter revenue of T$1.27 trillion ($39.6 billion), up 36% from a year earlier, beating analyst expectations. The surge is driven by insatiable demand for AI chips from customers including Nvidia and Apple. TSMC is scheduled to report full Q2 earnings Thursday and update its outlook for the rest of 2026.
ReutersAsian Tech Firms Raise Record $84B in 2026 as AI Demand Drives Unprecedented Capital Flows
Asian technology firms have raised a record $84 billion in the year to July 10, 2026 — more than triple the same period in 2025 — driven almost entirely by AI-related demand, according to LSEG data. SK Hynix's $26.5 billion U.S. IPO was the headline event, but analysts warn that subsequent Asian tech firms seeking U.S. listings may find foreign investors more selective unless they have similarly clear AI supply chain roles. SpaceX, meanwhile, has shed 35% from its post-IPO peak amid skepticism about its AI ambitions.
ReutersGeneral AI
2Federal Judge Warns: 'No Judge Should Be Abdicating Judicial Decision-Making to AI'
U.S. District Court Judge Jesse Furman, chair of the Judicial Conference's Rules of Evidence Advisory Committee, has publicly warned against judges delegating decision-making to AI systems, while acknowledging AI's potential utility for administrative and research tasks in the courts. Furman's comments come as AI use in legal contexts is rapidly expanding and courts are grappling with how to set boundaries. He emphasized that judicial reasoning must remain a human function.
Law.comUSPTO Launches AI-Based Prior Art Search Pilot — Patent Applicants Warned to Prepare
The US Patent and Trademark Office has announced an opt-in automated prior art search pilot program using AI tools, including custom-trained models and vector similarity search, to identify prior art references that could challenge patent claims. Bloomberg Law warns patent applicants to understand the long-term implications: AI-based search is more thorough and may surface prior art that traditional searches missed. The pilot signals a broader shift toward AI-assisted patent examination.
Bloomberg Law