A £35bn plan to fill the UK with 14 mini nuclear reactors
A Polish billionaire's company plans to build 14 small nuclear reactors across three UK sites for approximately £35 billion, targeting first power in 2034. The reactors will be GE Vernova Hitachi BWRX-300 units operated by Michał Sołowow's firm, SGE.
OpenAI launches GPT-Live, a full-duplex voice upgrade that lets ChatGPT talk more like a person
OpenAI launched GPT-Live, a full-duplex voice system that lets ChatGPT listen and speak simultaneously, replacing Advanced Voice Mode across iOS, Android, and ChatGPT.com. The two models, GPT-Live-1 and GPT-Live-1 mini, eliminate waiting for silence gaps by continuously processing incoming audio while generating responses.
A self-driving Waymo just reported its own passengers to police
A Waymo robotaxi detected two California teenagers drinking alcohol and shooting Orbeez from toy guns out windows, then parked itself and locked the passengers inside until police arrived. The vehicle's interior cameras and microphones, which monitor passenger behavior for safety and cleanliness purposes, enabled the detection and alert system.
The UN has an AI strategy for everyone except the labs that build it
The United Nations produced extensive AI governance frameworks across three years but focused entirely on downstream applications aligned with Sustainable Development Goals rather than the supply side. No multilateral body exists with technical staff to examine frontier AI laboratory work, evaluate training runs, or coordinate incident reporting across borders.
I tried Claude Cowork on my Gmail inbox after Gemini choked - and it saved me hours of work
Claude Cowork successfully identified relevant pitches, quotes, and permissions in a Gmail inbox where Gemini failed at the nuanced research task. The AI assistant demonstrated capability in handling complex email organization that saved significant time on what would otherwise require manual sorting and review.
Slack’s Slackbot can now pull your CRM data, generate charts, and send DocuSigns — all from a chat message.
Slackbot can now access Salesforce CRM data, Tableau analytics, and DocuSign through conversational prompts using dedicated Model Context Protocol servers. A salesperson can request customer deal history, view pipeline visualizations, update CRM records, and trigger document approvals without switching applications.
AI has collapsed the cyber response window — resilience now starts before the attack
Autonomous AI attacks can now breach systems in 27 seconds, faster than human security teams can detect and respond. Enterprises must shift focus to cyber resilience through identifying clean recovery states, mapping data dependencies, and automating restoration to recover systems within hours rather than days.
Tencent WeChat AI Agent Shows Promise in Super-App Fight: Review
Tencent developed Xiaowei, an AI agent prototype designed to run errands across WeChat's ecosystem of millions of apps for over a billion potential users. The company aims to restore its reputation as a tech innovator after lagging in frontier AI model development.
Samsung to Get Jump on Apple’s First Foldable With New Phones in July
Samsung will unveil the Galaxy Z Fold 8 on July 22 with a shorter and wider design. The new foldable phone resembles Apple's planned folding iPhone in its proportions.
An Xbox VP who spent 37 years at Microsoft was among those laid off yesterday
Microsoft eliminated 1,600 Xbox roles yesterday and plans to cut an additional 1,600 through the fiscal year, citing slower-than-expected growth in Game Pass and multiplatform content. Dave McCarthy, a corporate vice president with 37 years at Microsoft, left his position after eight years in his Xbox role.
The real cost, security, and culture problems behind enterprise AI agents
Enterprise AI agents operate at costs orders of magnitude higher than chatbots, shifting spending from engineering concerns to boardroom discussions. Organizations moving beyond pilot programs face security blind spots in autonomous systems and organizational friction that determines whether agent adoption spreads across the company.
New Zealand rules out a VPN ban after a fierce privacy backlash
New Zealand's government abandoned plans to restrict or ban VPNs after public backlash against its proposed under-16 social media ban. Prime Minister Christopher Luxon explicitly rejected the idea, stating there were no plans to ban VPNs at all.
Google, Meta, Spotify and Sony fight Belgium’s creator-pay law at the EU court
Belgium's copyright law requiring platforms to pay creators faces a legal challenge at the EU Court of Justice from Google, Meta, Spotify, and Sony. The companies argue the law imposes payment obligations that exceed EU copyright directive requirements.
Can AI fill prescriptions? Here’s what doctors think of Utah’s refill program
An AI chatbot called Doctronic launched in Utah this year to refill prescriptions online without doctor involvement, enabling residents to obtain refills through the program. The deployment challenges existing state and federal laws that restrict prescribing to licensed medical professionals, raising questions about AI regulation and safety measures in healthcare.
OpenAI is hiring an investment banker to teach its AI the job
OpenAI is hiring an investment banker earning $185,000 to $205,000 base salary to train its AI on investment banking practices. The position requires at least two years of investment experience and is located in San Francisco for the Applied AI team.
Chip Worker Shortfall Endangers US Factory Revival
A shortage of high-skilled workers threatens to delay construction of billions of dollars in new US semiconductor plants and constrain future chip production. The industry and government must coordinate resources and maintain funding to address the worker gap.
South Korea’s FuriosaAI brings its Nvidia-challenger chips to Europe
FuriosaAI activated its RNGD AI accelerators at Equinix's LS2 datacentre in Lisbon, marking the Korean chipmaker's entry into the European market. The startup positions its accelerators as a cooler and cheaper alternative to Nvidia's offerings for AI workloads.
Waymo’s robotaxis stalled, snarled traffic, and drove into fireworks on July 4
Waymo robotaxis experienced multiple failures during July 4th celebrations in San Francisco, including vehicles running out of charge, creating hours of traffic congestion, and one car driving over a lit firework that caused a fire. The incidents occurred across the holiday weekend as numerous driverless vehicles stalled in gridlock following the Golden Gate Bridge fireworks display.
Microsoft starts swapping OpenAI and Anthropic out for its own AI in some apps
Microsoft is replacing OpenAI and Anthropic AI models with its own in-house MAI models for certain product features based on cost and data residency considerations. OpenAI and Anthropic still handle most production traffic in Copilot, with the transition occurring incrementally rather than all at once.
AI learning loops aren’t an engineering trick. They’re a governance issue
AI systems are shifting from single prompt-response interactions to autonomous loops where agents continuously work, check, retry, and coordinate without waiting for human instructions. Unlike prompts that produce isolated outputs, loops create ongoing behaviors that observe, act, receive feedback, and adjust repeatedly, making errors compound rather than disappear.
Hacked, leaked, and held for ransom: The worst breaches of 2026 so far
A DOGE data breach, along with compromised energy, water, and FBI surveillance systems, represent the most damaging security incidents of 2026 so far. Critical infrastructure systems and government surveillance operations have been among the primary targets of hackers engaging in data theft and ransom demands.
Box survey: Why enterprise AI leaders are outperforming their peers
The share of organizations describing themselves as advanced or leading AI users surged from 8% to 64% over the past year, while early-stage organizations dropped from 53% to 9%. Eighty percent of organizations reported at least 10% return on their AI investment, with over half achieving measurable business impact within six months of project approval.
Anthropic brings Claude Cowork to mobile and web as usage data shows most users aren’t coding
Claude Cowork expanded to mobile and web platforms, allowing tasks to start on one device and continue autonomously in the background across devices. Usage data from 1.2 million sessions showed the majority of Cowork activity involved business processes and operations rather than coding tasks.
The foundational elements of AI architecture that IT leaders need to scale
Organizations are expanding AI use cases as capabilities advance and agentic systems develop, creating uncertainty about which investments will remain valuable. IT leaders must focus on foundational AI architecture elements to manage the risks introduced by rapid technological evolution.
Digital-native startups are ditching rigid databases for their agentic stacks
Digital-native startups are abandoning traditional relational databases for MongoDB Atlas to support AI agents that require variable schemas, vector embeddings, and real-time retrieval without manual migration management. Three startups, Huntr, Modelence, and Tavily, adopted this unified platform featuring native vector search and hybrid search capabilities to handle simultaneous operations at multi-tenant scale.
Chinese Firms Leave Nvidia for Local AI Suppliers, Survey Shows
Chinese companies are switching from Nvidia's advanced accelerators to domestic silicon suppliers due to US tensions. A survey shows this shift is reshaping AI infrastructure development and advancing China's goal of reducing dependence on American technology.
Judge upholds the verdict that Musk defrauded Twitter investors
A federal judge upheld a jury's finding that Elon Musk defrauded Twitter investors during his $44 billion takeover in 2022. The judge denied Musk's motion to overturn the verdict, which a San Francisco jury reached in March regarding two of his May 2022 tweets.
Utah let an AI renew prescriptions without a doctor, and its medical board wants it stopped
An AI chatbot in Utah began renewing prescriptions without doctor oversight starting in January, marking the first such program in the US. The state's medical board opposes the practice run by company Doctronic, which allows residents to refill prescriptions online through the chatbot.
An AI agent just ran a full ransomware attack with no human at the keyboard
An AI agent completed an entire ransomware attack autonomously, including reconnaissance, credential theft, and lateral movement across systems. Sysdig documented this first end-to-end AI-executed attack, which the firm named JadePuffer and involved a large language model planning and adapting each stage.