Archive

July 2026

Daily Brief

Daily Brief: robotics endpoints, model-platform exits and agent limits

Daily Brief — 31 July 2026 Google puts Gemini Robotics ER 2 into public preview Google has released two Gemini Robotics ER 2 endpoints in public preview: a general model and an on-device variant. Google says the release adds spatial reasoning, agentic code execution, multi-step tool orchestration, video moment finding, progress classification and multi-robot coordination. The practical change is that robotics teams can now evaluate those capabilities through the Gemini API rather than treat

2 min
Daily Brief

Daily Brief: GPT-5.6 gets cheaper as agents meet the real world

Lead — GPT-5.6’s lower price is a direct bid for production workloads. OpenAI has cut on-demand API pricing for GPT-5.6 Terra by 20% to $2 per million input tokens and $12 per million output tokens, while Luna falls 80% to $0.20 and $1.20 respectively. The change matters because long-running agent systems amplify token costs: a cheaper general-purpose model gives teams more room to keep context, retry tool calls and move routine automation out of prototypes. OpenAI is positioning the reductions

2 min
Daily Brief

AI Daily Brief: Security Moves From Model Demos to Infrastructure

Anthropic’s latest cryptography result is a useful warning: frontier models are starting to change the economics of finding flaws in security standards, not just bugs in implementations. Working with Claude Mythos Preview, the company found a stronger attack on HAWK, a post-quantum signature candidate, and improved attacks on a reduced-round AES variant. Anthropic says neither result affects deployed production systems, but its account puts the HAWK work at roughly 60 hours and $100,000 in API c

2 min
Daily Brief

Daily Brief: AI compute becomes infrastructure — and security goes open

The lead Meta brings BlackRock into the financing of its AI compute build-out Meta and BlackRock have formed a venture to develop and own a data-centre campus in El Paso, Texas. Meta frames the deal as a way to pair its operating expertise with outside infrastructure capital as it scales Meta Compute for its long-term AI ambitions. The concrete implication is financial as much as technical: the next wave of AI capacity is being structured as an infrastructure asset, letting a hyperscaler ad

3 min
Daily Brief

Daily Brief: Google maps AI use — and GitHub prepares MCP for statelessness

The lead Google’s ATLAS report puts numbers on how people use AI Google has released ATLAS, a large-scale, de-identified study built from 15 million interactions across the Gemini app, AI Mode and Gemini API. Its first finding cuts through some of the automation rhetoric: workplace use spans 68% of occupations representing 90% of US employment, but AI touches only about 21% of tasks in a typical job and fewer than 10% of work interactions fully automate a task. The dataset is not a census o

2 min
Wooden letter tiles spelling LAW on a table.
AI Regulation

Indonesia’s Draft Copyright Rules Put AI Training and Style Imitation on Notice

Indonesia is preparing a copyright-law rewrite that would explicitly address generative AI, including disclosure duties and a proposed bar on using AI to imitate a creator’s distinctive style. A draft reviewed by Reuters reporting would also require technology platforms to compensate publishers for aggregating, republishing or previewing news, and for using that material to train AI systems. The draft would recognise copyright in AI-assisted work where there is sufficient human involvement, wh

2 min
Enterprise AINews
Shanghai’s Pudong skyline at night.
AI Regulation

China’s New AI Alliance Is a Bid to Write the Rules Beyond Washington

China has launched the World Artificial Intelligence Cooperation Organisation, or WAICO, with 29 founding countries, turning a Shanghai AI conference into a concrete attempt to shape global AI governance. President Xi Jinping used the event to argue for international cooperation and wider access to AI capacity, as Reuters reported. The organisation is to be headquartered in Shanghai, placing its secretariat inside China’s commercial and policy centre. Al Jazeera’s account says the founding gro

2 min
China AINews
Server racks in a modern data centre.
AI Models

Moonshot’s Kimi K3 Makes the Open-Model Race a Scale Contest

Chinese startup Moonshot AI has unveiled Kimi K3, a 2.8-trillion-parameter system it calls the world’s largest open-weight AI model. The company says the model performs competitively with leading US systems, according to Reuters’ July 17 report. The important qualifier is open-weight, not merely large. Moonshot is promising developers and companies a model they can inspect, adapt and run through their own infrastructure rather than rent solely through an American lab’s API. The reported scale

2 min
China AINews
Urban apartment building facades beneath a clear blue sky
AI Regulation

New York City makes AI-altered rental listings a disclosure issue

New York City will require disclosure when rental listings have been altered with AI, folding synthetic imagery into a wider package of tenant-protection measures announced July 16. The Mayor's Office says the package contains 23 policy actions, including stronger enforcement against negligent landlords, oversight of fees and utility charges, and recognition of tenant unions. The AI measure is aimed at a familiar consumer harm: a renter can spend time and money visiting a property whose listin

2 min
News
Film crew working indoors with professional camera equipment
AI Models

Google puts a personal face on workplace AI video

Google is rolling out Gemini Omni video generation and personal avatars in Google Vids, turning a selfie and voice recording into a digital presenter for workplace videos. The company says the update, announced July 16, lets users generate or revise clips with text prompts and image references; Google's announcement says each AI-generated clip carries a digital watermark. The product sits inside Google Workspace's video tool, where the obvious use cases are training, internal announcements and

2 min
Enterprise AINews
Colourful programming code displayed on a computer monitor
AI Models

Thinking Machines Opens Its First Model—and Makes Customisation the Bet

Thinking Machines Lab has released Inkling, its first public AI model, as open weights under an Apache 2.0 licence. The company’s model card says the July 15 release accepts text, image and audio inputs and produces text; Reuters reported the release as a potential alternative to open offerings from Chinese labs. Inkling is a 975-billion-parameter multimodal model, according to reporting and the company’s release materials. It is not being positioned as the top benchmark model: Thinking Machin

2 min
Enterprise AINews
People using a smartphone at an outdoor event in China
AI Models

Apple Clears a Regulatory Gate for iPhone AI in China

Apple Intelligence has been registered with China’s cyberspace regulator for use on iPhones, removing a major regulatory obstacle to the service’s long-delayed Chinese launch. Reuters reported the registration on July 15; Apple still has to turn approval into an actual rollout. The clearance matters because Apple has been selling AI-enabled iPhones elsewhere while its flagship generative features have been absent from one of its most important markets. Reporting from Reuters and other outlets

2 min
AI RegulationChina AI
A row of hardcover patent books on a shelf
AI Models

Google faces fresh publisher challenge over Gemini training

Publishers and authors including Hachette, Cengage, Elsevier and Scott Turow have filed a class action accusing Google of using copyrighted books to train Gemini without permission. The TechCrunch reports on the new lawsuit was filed in the US District Court for the Southern District of New York and also alleges altered or removed copyright information. The plaintiffs say Google exceeded the limited permissions behind Google Books and Google Play by using copies for model training rather than

2 min
AI RegulationNews
Modern server hardware in a blue-lit data center
AI Infrastructure

New York freezes big AI data-center permits for a year

New York has ordered a one-year statewide halt to permits for proposed hyperscale data centers, becoming the first US state to impose such a pause. The Guardian reports covers facilities with more than 50 megawatts of electrical capacity and directs regulators to write standards for energy, water and environmental effects. The state says the freeze will remain while it considers rules that could make large operators provide their own power or pay more for electricity. Hochul framed the move as

2 min
AI RegulationNews
Technology workspace with a computer and desk tools
AI Infrastructure

Reflection locks in $1 billion of Nebius compute

Open-model startup Reflection AI has signed a $1 billion compute agreement with Nebius, giving it access to Nvidia's latest chips as it trains and serves models. TechCrunch reports that the deal follows a recent arrangement for SpaceX computing resources. Reflection, founded in 2024 by two former Google DeepMind researchers, is valued at $8 billion and has raised nearly $2.6 billion from backers including Nvidia, Sequoia and Lightspeed. Nebius is rapidly becoming a major supplier of scarce AI

2 min
AI ModelsFunding
Close-up of circuit boards in an electronics factory.
Funding

SK Hynix Turns the AI Chip Frenzy Into a $26.5 Billion Cash Call

SK Hynix priced its US listing on Friday to raise $26.5bn, turning investor appetite for AI infrastructure into one of the biggest stock sales on record, the Guardian reports. The chipmaker is selling about 18 million Nasdaq-listed depositary shares at $149 each after the deal was reportedly more than seven times oversubscribed. The memory specialist has become one of the clearest financial winners from the datacenter build-out because it supplies high-bandwidth memory used alongside Nvidia ac

1 min
HardwareNvidia
Developers working together on computers in a modern tech office.
AI Models

OpenAI’s GPT-5.6 Push Is Really a Pricing War for Enterprise AI

OpenAI rolled out GPT-5.6 on Thursday as a three-model family aimed squarely at enterprise buyers and coding teams, TechCrunch reports. The lineup splits into Sol, Terra and Luna, with OpenAI pitching Sol as its top coding system and saying it uses 54% fewer tokens on coding tasks than prior models. TechCrunch says OpenAI is also shipping a companion product called ChatGPT Work for document, spreadsheet and presentation tasks across desktop, web and mobile. The company is framing GPT-5.6 as it

1 min
Enterprise AINews
A modern server unit in a blue-lit data center.
AI Infrastructure

Ed Markey Wants Datacenters Treated Like Public-Interest Infrastructure

US senator Ed Markey used a new "AI accountability agenda" on Friday to bundle bills aimed at datacenters, hiring algorithms, child-safety failures and discriminatory automated systems, the Guardian reports. The center of gravity is a draft proposal that would force AI datacenters to secure federal certification before construction. Markey argues the boom is pushing environmental and social costs on to everyone else. His preliminary text would have the FCC weigh air and water quality, noise, e

1 min
AI RegulationNews
A neoclassical government building with an American flag and cherry blossoms in Washington, DC.
Hardware

Apple’s Hardware Counterpunch to OpenAI Gets Messy Fast

Apple sued OpenAI on Friday, alleging the ChatGPT maker raided its hardware operation for confidential product information as it pushed into devices, the Guardian reports. Apple says former staff were persuaded to share unreleased designs, supplier details and other internal material tied to future products. The complaint names OpenAI hardware chief Tang Yew Tan, formerly an Apple vice-president, and another ex-Apple employee, Chang Liu. Apple says Tan pushed candidates to bring real component

1 min
NewsOpenAI
Two professionals in a high-tech environment working with cash and computers, analyzing complex data.

Google's SynthID Watermark Exposes AI-Generated McConnell Hospital Hoax

Google's SynthID system scored a rare public win this week, TechCrunch reports, after Snopes used it to debunk a widely-shared AI-generated image of Senator Mitch McConnell. The fake showed the 84-year-old Kentucky Republican covered in tubes in a hospital bed in extreme distress, fuelling speculation about his health after a June 14 emergency hospitalisation. SynthID embeds an invisible signature into AI-generated images that survives screenshots and cross-platform sharing — the watermark was

1 min
Close-up of a modern server unit in a blue-lit data center environment.

Prime Intellect Hits $1B Valuation With $130M Raise for DIY Enterprise AI Agents

Prime Intellect, a startup helping companies build their own AI agents without relying on OpenAI or Anthropic, has raised $130 million in Series A funding at a $1 billion valuation, TechCrunch reports. Radical Ventures led the round, joined by Nvidia Ventures, Intel Capital, and Dell Technologies Capital. Founded in 2024, Prime Intellect offers a "full stack" agent development platform — compute, reinforcement learning frameworks, and evaluation tools — sold as modular building blocks. Custome

2 min
Close-up of a smart speaker with digital time display indoors.

OpenAI Launches GPT-Live-1 Voice Models That Listen and Speak at the Same Time

OpenAI released two new conversational models on Tuesday, TechCrunch reports — GPT-Live-1 and GPT-Live-1 mini — that can speak and listen simultaneously, letting users interrupt naturally and enabling live translation. ChatGPT's Advanced Voice Mode switches to GPT-Live-1 mini by default; paid users unlock the full model. The new system fixes longstanding complaints: the old voice mode would cut people off mid-sentence and lacked the intelligence for complex questions. Now queries route to GPT-

2 min
A rocket launch with smoke trails over South Padre Island, viewed across the ocean.

SpaceXAI Ships Grok 4.5, Musk Calls It an 'Opus-Class Model' at Half the Price

SpaceXAI released Grok 4.5 on Wednesday, TechCrunch reports, the company's first model since going public weeks ago. Elon Musk described it as an "Opus-class model" — a direct shot at Anthropic's premium tier — claiming it's "faster, more token-efficient, and lower cost." Grok 4.5 costs $2 per million input tokens and $6 per million output tokens. Anthropic's Opus 4.7 runs $5 and $25 respectively. OpenAI's top-tier Sol sits at $5 and $30. SpaceXAI also claims "twice greater token efficiency" t

1 min
Server rack with glowing lights in a modern data center

Scotland Moves to Freeze All New Datacentre Projects, Threatening UK AI Strategy

The Scottish government is set to consider a sweeping moratorium on new datacentre construction after the SNP's national council passed a motion to freeze all projects not yet granted planning permission. As the Guardian reports, 24 hyperscale datacentre projects are currently planned across Scotland, collectively requiring more than one-and-a-half times the country's peak electricity demand. The move threatens a central plank of the UK's AI infrastructure strategy, which has positioned Scotla

1 min
Smartphone screen displaying popular social media applications

Reddit Deploys LLMs to Fight LLM-Generated Spam, Blocking 23 Million Views Daily

Reddit is using large language models to combat spam created by large language models, the company disclosed in a blog post. The platform now blocks 23 million spam views per day and catches approximately 25,000 new spam posts and comments daily. Reddit says LLM-powered detection catches "highly subtle, coordinated patterns of fake behavior and artificial hype that older systems once missed." User exposure to spam dropped 20% from January to March compared with the prior three months. TikTok n

1 min
Server racks in a data center highlighting modern technology infrastructure

Vercel Pushes to Split AI Models From Agents as Deployments Hit 6 Million a Day

Vercel CEO Guillermo Rauch has called for a structural separation between AI models and the agents that use them, warning that developer tools training on proprietary codebases pose an existential data risk. Speaking to TechCrunch after the company's ShipNYC conference, Rauch revealed Vercel now handles 6 million deployments daily — half triggered by coding agents — with more than 1 trillion tokens flowing through its AI gateway. Vercel's new Eve framework lays out agent instructions in natura

1 min
United Nations Office in Geneva with flags of various countries

UN Panel Warns Window for Global AI Governance Is Closing

A UN scientific panel has warned that the window for effective global AI governance "remains open but may not stay that way for long," as the Global Dialogue on AI Governance began Monday in Geneva. The preliminary report by the UN Independent International Scientific Panel on AI, launched on 1 July, finds AI capabilities outstripping regulatory capacity at an accelerating rate. The report notes that AI agents — systems that plan tasks and complete complex assignments with little human oversig

1 min
Rocket launch

Grok 4.5 Enters Private Beta at SpaceX and Tesla With 1.5 Trillion Parameters

Elon Musk announced on June 28 that Grok 4.5 has entered private beta at SpaceX and Tesla, running on a 1.5 trillion parameter V9 foundation model — a 50% scale increase from Grok 4.4 in roughly one month. "Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX and Tesla," Musk posted on X. "Early evals show performance close to, perhaps exceeding Opus." The Cursor training data comes from SpaceX's $60 billion acquis

2 min
Technology in China

Alibaba Bans Employees from Using Anthropic's Claude Code

Chinese e-commerce giant Alibaba will ban employees from using Anthropic's programming tool Claude Code starting July 10, TechCrunch reports, classifying the tool as high-risk software and directing staff to use the company's own Qoder tool instead. The ban follows Anthropic's escalating efforts to block Chinese access to its models. The company already prohibits Chinese firms and their foreign subsidiaries from using Claude, and has been closing loopholes that allowed users in China to connec

2 min
Documents and paperwork on a desk

White House Finalising Voluntary AI Release Standards, Announcement Expected Within Days

The White House is in advanced talks with OpenAI, Google, and Anthropic to finalise voluntary standards for frontier AI model releases, the Financial Times confirmed, with an announcement possible as soon as next week. Technical teams from the three AI labs have met consistently with White House officials to negotiate two sticking points: how long the government can hold a model before public release, and what capability threshold triggers a review. The framework would replace what former Whit

2 min
US Capitol building

OpenAI Proposes Handing US Government a 5% Stake Ahead of IPO

OpenAI has proposed giving the United States government a 5% equity stake in the company in an extraordinary move, the Financial Times reports, representing what would be the largest voluntary equity transfer from a private tech company to the US government in history. The stake, worth an estimated $15 billion at OpenAI's current $300 billion valuation, comes as the company races toward an IPO targeting $830 billion to $1 trillion. CNBC also confirmed the talks, tying them to Trump administrat

2 min
Artificial intelligence technology visualization

UK to Ban Under-16s from Social Media, Opens AI Growth Lab for Legal Sector

The UK government has announced plans to ban under-16s from social media and launched an AI Growth Lab targeting the legal sector, as part of a policy package published alongside a major foresight report mapping five possible AI futures by 2030, TLT reports. The social media ban mirrors measures adopted in Australia and debated across Europe, positioning the UK among the most aggressive jurisdictions on child online safety. The AI Growth Lab is aimed specifically at legal services, a sector th

2 min
White House exterior

White House in Advanced Talks for Voluntary AI Model Release Standards

The US government is in advanced negotiations with AI companies to establish voluntary standards for releasing new frontier models, with an announcement possible within the next week, the Financial Times reported on Wednesday. The framework would let officials designate a model as a "covered frontier model" and negotiate access terms before it ships, without imposing a licensing requirement the White House has explicitly ruled out. The talks build on Trump's June executive order, which asked d

2 min
Stock market display with financial data

Alphabet Jumps 4% on Dow Debut as AI Spending Drains Cash and Talent

Alphabet shares rose 4% Monday on the company's first day in the Dow Jones Industrial Average, replacing Verizon — but the ceremonial win masks deeper trouble. The stock is on track for its worst month since February 2024. Alphabet has raised more than $140 billion in debt and equity to fund AI infrastructure. In the first quarter, the company skipped share buybacks for the first time in roughly a decade. Cash reserves are shrinking as it pours money into computing capacity it still does not h

2 min
UN Scientific Panel Warns AI Governance Window Is Closing Fast

UN Scientific Panel Warns AI Governance Window Is Closing Fast

A preliminary report by the UN Independent International Scientific Panel on Artificial Intelligence warns that the window to establish effective global AI governance "may not stay open for long," as the complexity of tasks AI systems can complete doubles every few months. Launched Wednesday, the report documents a dramatic acceleration: AI has moved from answering questions and generating text to writing code, discovering medicines, and operating AI "agents" that plan tasks, use digital tools

2 min
SpaceX Shows Investors an AI Handset That's Slimmer Than an iPhone

SpaceX Shows Investors an AI Handset That's Slimmer Than an iPhone

Elon Musk's SpaceX has shown investors a prototype of a "handset-like" AI device, reportedly sleeker than an iPhone and built to run a proprietary operating system with technology from xAI, The Wall Street Journal reported, citing people familiar with the demonstrations. The device was shown to investors and stakeholders before SpaceX went public. Sources said the design is at an early stage and could still change. Musk has denied the reporting, calling it "utterly false" on X. SpaceX and si

2 min
OpenAI Offers Trump Administration 5% Stake Worth $42.6 Billion

OpenAI Offers Trump Administration 5% Stake Worth $42.6 Billion

OpenAI is discussing handing the Trump administration a 5% equity stake in the company, a share worth roughly $42.6 billion based on its March funding valuation of $852 billion, the Financial Times reported Thursday, citing people familiar with what were described as "early conversations." CEO Sam Altman has argued the arrangement would let the public share in AI's financial upside, framing it as a sovereign-style wealth fund similar to the Alaska Permanent Fund, which invests state oil revenu

2 min
CIA Director Calls AI a 'Digital Nuclear Weapon' as Agency Restructures for the Age of AI

CIA Director Calls AI a 'Digital Nuclear Weapon' as Agency Restructures for the Age of AI

CIA Director John Ratcliffe announced a "fundamental reshaping" of the agency's entire approach to technology on Monday, calling artificial intelligence a "digital nuclear weapon" and revealing that the CIA has signed nearly 400 technology contracts in the past six months alone. The restructuring merges the CIA's technology and acquisition offices into a single directorate, designed to cut the time between identifying a technology need and deploying a working system. Ratcliffe said the rapid a

2 min
Etched Exits Stealth at $5 Billion Valuation With $1 Billion in AI Chip Contracts

Etched Exits Stealth at $5 Billion Valuation With $1 Billion in AI Chip Contracts

AI chip startup Etched emerged from stealth on Monday with a working chip manufactured by TSMC, $800 million in funding raised, a $5 billion valuation, and over $1 billion in signed customer contracts for its Sohu inference processor. Founded by two Harvard dropouts and based in San Jose, Etched is betting that the future of AI hardware belongs to application-specific integrated circuits, not general-purpose GPUs. Its Sohu chip is an ASIC built exclusively for transformer model inference — doi

2 min
Anthropic Launches Claude Sonnet 5 as Export Controls on Fable 5 Are Lifted

Anthropic Launches Claude Sonnet 5 as Export Controls on Fable 5 Are Lifted

Anthropic shipped three major updates on Monday: Claude Sonnet 5 became the new default model for Free and Pro users, Claude Science launched as a research workbench for scientists, and the US Commerce Department lifted export controls on Fable 5 and Mythos 5. Sonnet 5 is priced aggressively at $2 per million input tokens and $10 per million output tokens through August 31, positioning it as a cheaper workhorse for coding, research, and long-running agentic tasks. Anthropic says the model outp

2 min