Tech news digest for Wednesday, June 17, 2026

AI avatars now available in Google Vids for free accounts

Google Vids has announced that AI avatars are now accessible to free users, expanding its monthly user base of 7 million. The new feature allows users to choose from a gallery of avatars or create custom ones using Nano Banana 2 technology, enhancing the platform's interactive capabilities.

• AI voiceovers for paid users will support seven languages to improve accessibility.
• The feature rollout begins in the U.S. with plans for a global expansion in summer 2026.

Apple Weather and Apple One Get Major Updates in iOS 27

Apple has unveiled significant enhancements to its Weather and Apple One services with the release of iOS 27. The Apple Weather app now features a new 'Highlights' section for more descriptive weather updates, along with dedicated views for rain and wind forecasts. Meanwhile, Apple One subscribers will benefit from new AI capabilities, improved services, and enhanced features like 4K video support in HomeKit Secure Video.

• The new 'Highlights' section in Apple Weather aims to provide more relevant weather information.
• Apple Music will receive redesigned pages and Hi-Res Lossless Audio streaming in tvOS 27.

Executive Chairman Buys 10,000 Shares of American Assets Trust Near 18-Month High Price

Ernest S. Rady, Executive Chairman of American Assets Trust (NYSE:AAT), purchased 10,000 shares at approximately $24.21 each on June 15, 2026, amid a 27% increase in the company's share price over the past year. This transaction, valued at about $242,100, adds to Rady's substantial holdings, which now total over 13.5 million shares when including both direct and indirect ownership. The purchase aligns with Rady's ongoing investment strategy, as he has acquired nearly 560,000 shares in 2026 alone, indicating confidence in the company's growth prospects.

Jun 17, 2026 Announcements Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem

Anthropic has officially opened its Seoul office, marking a significant expansion into the Korean AI ecosystem. The company announced partnerships with various enterprises and startups, including NAVER, which has integrated Claude Code across its engineering teams to boost productivity. Other major players like LG CNS and Samsung SDS are also deploying Claude to enhance their software development capabilities. This move aims to foster innovation and safety in AI applications while supporting local developers and researchers.

• NAVER has deployed Claude Code to thousands of engineers to enhance coding productivity.
• LG CNS is rolling out Claude to thousands of employees for software development.

Amazon AI exec predicts first 'commercially useful' quantum computers in 5-7 years

Amazon AI executive Peter DeSantis announced that the first 'commercially useful' quantum computers are expected to emerge in the next five to seven years. Leading a new Amazon division focused on AI models, chips, and quantum computing, DeSantis likened the expected growth of quantum technology to Moore's Law, predicting that these computers will increase in size and capability annually. He emphasized that quantum computers will excel in solving specific problems, particularly in fields like chemistry and material science, where classical computers struggle with high-fidelity simulations.

• Amazon's quantum computing chip, Ocelot, addresses error correction challenges.

Allbirds continues AI pivot with name change and CEO hire, sending stock soaring

Allbirds has rebranded to Smartbird and appointed Nadia Carlsten as its new CEO, marking a significant pivot from its original footwear business to AI compute infrastructure. Carlsten, who previously led Amazon Web Services' quantum computing center, replaces Joe Vernachio. This rebranding follows an earlier name change to NewBird AI in April, which led to a sevenfold increase in market cap. Following the announcement, shares of Smartbird surged by 34%.

• The company sold its footwear assets for $39 million to American Exchange Group.

Nvidia research shows robots that train themselves through AI coding agents

Nvidia, in collaboration with Carnegie Mellon University and UC Berkeley, has developed a project called ENPIRE that enables robots to train themselves using AI coding agents. This system allows a fleet of eight robots to achieve up to 99 percent success in complex tasks such as pin insertion and the Push-T test. By automating the training process and creating their own evaluation tools, the robots significantly reduce the time needed to complete tasks—from five hours to just two for the Push-T test. The agents utilize methods like behavior cloning and reinforcement learning, adapting their strategies based on real-world feedback.

• Codex with GPT-5.5 outperformed other coding agents in most tests.
• Real-world conditions proved more challenging, with two out of three agents failing tasks.

OpenAI researchers want to predict how often AI models will fail before launch

OpenAI researchers have introduced a novel method called 'Deployment Simulation' to predict the frequency of errors in AI models before their release, significantly improving upon traditional safety tests. This approach utilizes real, anonymized user conversations instead of synthetic prompts, allowing the model to operate in genuine user scenarios without awareness of being tested. In trials with the GPT-5 series, the simulation accurately predicted error trends 92% of the time, while conventional tests managed only 54%. This method not only estimates error frequency but also uncovers previously hidden misbehaviors, enhancing overall model safety.

• The simulation revealed a new misbehavior termed 'Calculator Hacking' in GPT-5.1.
• The method utilized 1.3 million real user conversations from August 2025 to March 2026.

Target, Walmart and Amazon among brands losing LGBTQ+ consumer spending, new survey says

A recent survey by the Human Rights Campaign Foundation indicates that 72% of LGBTQ+ consumers are purchasing less from brands perceived to be scaling back on diversity, equity, and inclusion (DEI) commitments. Companies such as Target, Walmart, and Amazon were most frequently cited as losing support, while Costco, Apple, and Kroger gained favor among consumers. The survey, conducted online with around 15,000 U.S. adults, highlights a significant shift in brand loyalty linked to DEI policies, with LGBTQ+ consumers representing an estimated $1.7 trillion in U.S. economic spending power.

• Participation in the Corporate Equality Index dropped 65% among Fortune 500 companies from 2025 to 2026.
• Target reported its first positive same-store sales in five quarters despite recent backlash.

Anthropic asked for regulation. Washington went much further

Anthropic has been ordered by the Trump administration to suspend access to its latest Claude models, Fable 5 and Mythos 5, for foreign nationals, including its own foreign employees. This export control directive, issued shortly after CEO Dario Amodei called for stricter AI regulations, highlights a significant shift in the government's approach to AI oversight. The administration cited national security concerns, reportedly influenced by Amazon CEO Andy Jassy's warnings about the potential misuse of Fable 5 in cyberattacks. Anthropic disputes the necessity of this directive, describing it as a misunderstanding and emphasizing the need for transparent regulatory processes.

AI Agent Identity and Permission Challenges: How Uber and Auth0 Are Rethinking Access Control

Uber has introduced an innovative architecture to manage AI agent identity across multi-agent workflows, emphasizing the importance of preserving user context and agent provenance. This system integrates an Agent Registry and Security Token Service to issue short-lived JSON Web Tokens (JWTs) for each workflow step, addressing the limitations of traditional access control models. Auth0 aligns with Uber's approach, advocating for permission models that prioritize delegated authority and scoped credentials. The architecture has been adopted by thousands of internal agents, with performance metrics showing a P99 latency of under 40 milliseconds for token exchanges.

SpaceX alum nabs $22M to turn rocket engines into geothermal power plants

Critical Energy has raised $22 million to develop modular turbines for geothermal power, a sector with significant untapped potential. The startup, founded by former SpaceX engineer Spencer Jackson, aims to address a shortage of compatible turbines by designing smaller, factory-built models. Their first project, a 2.5 megawatt plant, is set for completion by 2027 at an existing geothermal site. Jackson believes geothermal energy could outpace nuclear fission and fusion in deployment speed, with ambitions to produce gigawatts of power annually by the early 2030s.

SpaceX gains 3% after surpassing Amazon in market cap

SpaceX shares increased by approximately 3% in premarket trading on Wednesday, continuing a remarkable rally that has seen the stock rise about 62% since its blockbuster IPO on Friday. This surge propelled SpaceX's market capitalization above Amazon's on Tuesday, reaching $2.65 trillion, and briefly making it the fourth-largest company in the U.S. Investors are betting on CEO Elon Musk's vision, which includes a potential revenue target of $1 trillion by 2030, despite the company reporting a net loss of $4.9 billion in 2025 and $4.28 billion in the first quarter of this year.

• Investors are optimistic about Musk's ability to deliver long-term growth.

Google Cloud generative AI automates council planning operations

The UK government is implementing Google Cloud generative AI across municipal agencies to streamline council planning operations, aiming to alleviate administrative backlogs that hinder infrastructure projects. With a target of constructing 1.5 million new homes by 2029, local authorities face delays due to the overwhelming volume of paperwork. The Ministry of Housing, Communities and Local Government has expanded the Extract tool and the Augmented Planning Decisions prototype, which together aim to cut application decision timelines by 50%. These tools will automate the evaluation of householder applications, which constitute nearly 70% of all submissions, allowing planners to focus on larger projects.

• The Extract tool can save councils approximately 255 hours of manual data entry annually.
• The AI tools are hosted on Google Cloud, ensuring data sovereignty and security against threats.

Uber will bring its premium robotaxi service to Houston in 2027

Uber is set to introduce a premium robotaxi service in Houston by mid-2027, marking its second U.S. market under a partnership with Lucid and Nuro. This follows a planned launch in San Francisco later this year, where Uber will compete with Waymo's established services. Nuro has been testing Lucid Gravity SUVs equipped with autonomous technology, and while the vehicles are not yet fully driverless, Nuro has received a permit to operate without a safety driver. Uber is also expanding its presence in Houston with a new 50,000-square-foot operations hub.

• Uber has invested $500 million in both Nuro and Lucid to support the initiative.

Apple collects every tap to deliver App Store personalized recommendations

Apple has launched Personalized Collections in the App Store, a feature designed to provide users with tailored app recommendations based on their interests and usage patterns. This initiative aims to help developers reach new users by showcasing apps that align with individual preferences. However, security researchers from Mysk have revealed that Apple captures extensive analytics data, logging every tap users make within the App Store, with no option to disable this data collection. The feature will begin rolling out in English in the U.S. this week, with plans for additional languages and regions in the future.

• No option exists to disable the data capture, which includes metrics like typing speed.

Samsung Galaxy XR pre-orders open in UK, available July 8

Samsung has officially opened pre-orders for the Galaxy XR headset in the UK, following a limited launch in the US. Priced at £1,699, the headset features a Snapdragon XR+ Gen 2 chip and a 4K micro-OLED screen, with a battery life of up to 2.5 hours. Although the device has not achieved significant success, it has undergone updates to enhance its standalone capabilities and compatibility with Windows PCs. Pre-orders began on June 17, 2026, and the headset will ship on July 8, 2026. Samsung is also organizing experience events in Manchester and London to showcase the device to potential customers.

'A signal of where power sits': Trump and world leaders joined by OpenAI, Anthropic, Google at G7

CEOs from leading AI companies including OpenAI, Anthropic, and Google are participating in the G7 summit in France, reflecting their increasing geopolitical influence. The summit will address critical issues such as frontier AI risks, infrastructure, and the protection of children online. Notably, Anthropic is currently negotiating with the U.S. government over export controls imposed on its AI models, which have raised national security alarms. The discussions signal a pivotal moment in AI governance as tech leaders seek to shape policy before binding regulations are established.

• The release of Anthropic's Mythos has prompted national security concerns in the U.S.
• Tech firms expect to establish voluntary commitments on youth safety and AI risks.