Jeff Bezos has revealed details about his new AI startup, Prometheus. This ambitious venture aims to create an "artificial general engineer." This AI will be capable of assisting in the design of complex physical products. Examples of such products include robots, drugs, manufacturing systems, and rocket engines. Prometheus recently secured $12 billion in funding, valuing the company at $41 billion. Bezos is serving as co-CEO alongside Vik Bajaj, a former leader at Alphabet's Verily. The startup currently employs approximately 150 individuals. The tools developed by Prometheus hold the potential to accelerate innovation across various industrial sectors. Bezos specifically highlighted how companies like his aerospace firm, Blue Origin, could greatly benefit from these AI-powered design capabilities.

The Justice Department has approved Paramount Skydance's acquisition of Warner Bros. Discovery, removing a significant hurdle for the $111 billion deal. This approval came without requiring the companies to sell any assets or make other concessions. The Justice Department concluded that the transaction would not harm competition. However, the deal still faces scrutiny from state attorneys general, with California's Attorney General actively investigating. The proposed merger aims to combine Paramount with the entertainment company that owns HBO Max, CNN, and a large film and television studio. This would create a new offering with approximately 200 million subscribers by merging HBO Max with Paramount+. Industry insiders and workers express concerns about potential mass layoffs and reduced opportunities for creators. Paramount, however, argues the merger will enhance competition against streaming giants and tech companies. Despite the DOJ's go-ahead, the potential for legal challenges from the state level remains. The entertainment industry is contemplating the significant reshaping of Hollywood due to this consolidation.

The hacking group ShinyHunters has claimed to have compromised over 100 organizations using a zero-day vulnerability in Oracle PeopleSoft. One confirmed victim is the University of Nottingham, from which ShinyHunters claims to have stolen 40GB of student and billing data. The group posted the university on its data leak site and later published the stolen files, reportedly after the university refused to pay an extortion demand. ShinyHunters stated they have begun contacting other affected organizations to reach an agreement. A Google threat intelligence report corroborated these claims, identifying malicious activity consistent with exploiting CVE-2026-35273 between late May and early June. Google notified over 100 global organizations whose IP addresses were associated with potentially vulnerable endpoints. The majority of these affected organizations are located in the US, with 68 percent belonging to the higher education sector. Oracle has released a patch availability document, but it is uncertain if a patch is presently accessible. The full extent of ShinyHunters' other victims and their planned data leak timeline remains undisclosed. This incident highlights a significant security breach impacting numerous institutions globally.

Google is suing to dismantle Outsider Enterprise, an alleged Chinese cybercrime network. This group is accused of using AI to send vast numbers of scam text messages. These messages impersonate Google and other brands to steal sensitive information. Outsider Enterprise has reportedly scammed hundreds of thousands of victims, losing millions. In just two weeks, they deployed 9,000 fake websites and 1 million fraudulent domains. They also sent 2.5 million texts to Android users during that period. Google noted over 55,000 spam texts were flagged by users in May. The tech giant uses its own AI tools to combat AI-powered scams and protect users. Google intercepts over 10 billion scam messages monthly through these tools. They are collaborating with major carriers and the FBI to block messages and take legal action.

A leaker with a strong Apple rumor track record claims a touchscreen MacBook is 100% confirmed. This would represent a significant shift for Apple, which has historically opposed touchscreens on Macs. Several recent reports corroborate this expectation of a touchscreen MacBook. This feature is anticipated to be part of major upgrades for upcoming high-end MacBook Pro models. Other rumored enhancements include M6 Pro and M6 Max chips, an OLED display, and the Dynamic Island. The new laptops might even be branded as MacBook Ultra. macOS 27 also includes a more touch-friendly interface, enhancing interaction via iPad's Sidecar. Apple is expected to market these as "touch-friendly" rather than "touch-first" devices. Users will reportedly be able to use touch and mouse gestures interchangeably for all functions. This development potentially revisits the debate around touchscreen laptop functionality.

A firmware flaw in Microsoft Surface devices could brick them with a single packet, but only if Secure Core and Secure Boot were disabled. Microsoft Copilot inadvertently helped identify this vulnerability when researchers asked it to adjust screen backlighting. The AI generated Python scripts, which overwrote embedded controller firmware, rendering a laptop inoperable. Microsoft confirmed the issue, stating a deprecated UEFI interface could trigger a boot loop under specific user configurations. They have released updates to address the problem for most affected devices. However, users running Linux, or those who have disabled security features on Windows, may still be vulnerable if their systems haven't been patched. The vulnerability appears to affect various Surface Laptop and Book models, excluding Surface Go devices, though ARM variants were not tested. Microsoft is actively working to enhance Surface device security by migrating components to Rust. This includes rebuilding embedded controller firmware and UEFI DXE Core in Rust for improved reliability and security. These initiatives, known as Secure EC and Project Patina, are open-source to promote transparency and broader ecosystem adoption.

Sam Bankman-Fried's appeal to overturn his FTX fraud conviction and 25-year sentence has been denied. A three-judge panel of the 2nd U.S. Circuit Court of Appeals unanimously upheld the conviction. The court found that the evidence presented by prosecutors was robust. They stated that Bankman-Fried misused FTX customer funds for personal expenses while assuring customers their money was safe. His lawyers have not yet commented on the decision. They may pursue further appeals, potentially with all active judges of the 2nd Circuit or the U.S. Supreme Court. Additionally, Bankman-Fried is reportedly seeking a presidential pardon from Donald Trump. He was sentenced to 25 years in prison in 2024 for orchestrating a massive financial fraud. Bankman-Fried was convicted on multiple charges, including wire fraud, conspiracy, and money laundering. The judge described his actions as using FTX as a personal piggy bank.

Infineon is opening a $5.8 billion power-chip fabrication plant in Dresden on July 2, supported by approximately $1.1 billion from EU Chips Act subsidies. This new facility will manufacture power semiconductors essential for AI data centers. The growing demand for AI infrastructure is significantly impacting global electricity systems, as AI data centers are projected to double their electricity consumption by 2030. Infineon, historically focused on automotive chips, is now benefiting from the surge in demand for power chips used in these data centers. The Dresden fab's production will ramp up based on demand, potentially generating up to 5 billion euros in annual revenue. Infineon has invested around 2 billion euros in construction, with future spending allocated for additional machinery. Analysts view the new facility as a crucial driver for Infineon, noting that AI customer demand currently exceeds the company's production capacity. This imbalance is expected to improve in the 2027 and 2028 fiscal years, with a raised revenue forecast for AI power chips. Infineon anticipates data center-related revenue to increase from roughly 1.5 billion euros in fiscal 2026 to 2.5 billion euros in 2027. While Infineon does not produce advanced AI chips like those from Nvidia, its power semiconductors are critical components for AI infrastructure.

Elon Musk, a deeply embedded figure in popular culture, has achieved the unprecedented status of the world's first trillionaire. Despite widespread concerns about inequality and negative views of the ultra-wealthy, Musk maintains a loyal following without adopting a folksy persona like Warren Buffett. His unfiltered style appeals to admirers, though critics accuse him of oligarch-like power, corporate governance issues, and partisan political interventions. Nevertheless, investor enthusiasm for his ventures remains high. SpaceX, Musk's rocket, satellite, and AI company, alongside electric-car maker Tesla, forms the core of his empire. SpaceX recently raised a record $75 billion in its initial public offering. Prior to this sale, Forbes estimated his net worth at $780 billion, significantly surpassing the next wealthiest individual, Larry Page. Matt Durot of Forbes Wealth noted the substantial gap, highlighting Musk's unique financial position. Most of Musk's wealth is now tied to his roughly $866 billion stake in SpaceX. With the addition of Tesla and his other assets, his net worth is projected to exceed $1.1 trillion once SpaceX stock begins trading.

Pokemon Go players' voluntary location scans reportedly helped train Niantic Spatial's visual positioning system. This system allows machines to navigate using camera imagery and 3D maps when GPS is unavailable. Niantic Spatial's technology is now being integrated with Vantor's drone navigation software. This partnership aims to support military and intelligence operations in GPS-denied environments. The process involved players scanning the physical world, which Niantic Spatial converted into 3D maps for machine self-location. The major concern is whether players understood their data could contribute to defense systems. The author highlights the global issue of drones losing navigation when GPS is jammed. While the technology itself isn't the unsettling part, the origin and intended use of the training data are. A Niantic spokesperson stated that Pokemon Go data is no longer shared with Niantic Spatial since its acquisition by Scopely. They also noted that AR Scans were voluntary and subject to terms of service at the time. The company discontinued AR scanning and data sharing as part of Pokemon Go's transition to Scopely.

Riverside County has initiated an eight-mile "smart freeway" pilot on northbound I-15 near Temecula. This $33 million project utilizes roadway sensors and an algorithm to manage ramp meters and suggest speeds. The goal is to reduce stop-and-go traffic and travel times without widening the freeway. Similar systems in Australia and Denver have reportedly cut delays by 20% to 65%. Drivers on I-15 might experience ramp waits of up to four minutes as the system optimizes traffic flow. Transportation officials aim to improve freeway flow and decrease travel times by spacing out vehicles. This two-year pilot program's success could lead to its expansion countywide and potentially across California. The system is managed by advanced sensors monitoring real-time traffic, not artificial intelligence. The chosen stretch of I-15 is known for significant congestion, especially during peak hours. The intent is to create a consistent traffic flow by coordinating ramp metering to prevent frustrating stop-and-go conditions.

China is emerging as a significant destination for medical tourism, differentiating itself from traditional Asian hotspots. Unlike countries focusing on cosmetic surgery or general health checks, China is attracting patients with advanced medical treatments. This trend is driven by both the availability of cutting-edge procedures and more affordable pricing. While simpler treatments like acupuncture are gaining popularity on social media, CAR-T cell therapy is a key advanced treatment drawing international attention. CAR-T therapy, a complex cancer treatment, is prohibitively expensive in countries like the US, but significantly cheaper in China. China's medical tourism market is still developing, with pilot zones like Hainan slowly increasing foreign patient numbers. However, the country is actively promoting its high-value healthcare services to diversify its economy and improve its global image. The global medical tourism market is expected to grow substantially, with China's sector projected for significant expansion. Patients are increasingly choosing China for treatments unavailable or too costly elsewhere, attracted by advancements in medical technology and cost-effectiveness. Integrating these advanced treatments into the international system is seen as mutually beneficial for China and global healthcare access.

Two recent studies suggest smartphones may be a factor in declining birthrates. One study, "The Collapse of Teen Fertility in the Digital Era," posits that teen fertility globally began to decline around 2007, coinciding with the iPhone's release. It argues that smartphones altered how teens interact, shifting peer networks online and reducing in-person time. This decreased unstructured contact, where most unintended teen conceptions occur, leading to a sharp drop in teen fertility. The study also noted a concurrent rise in teen suicides, suggesting the phone-mediated shift affects multiple aspects of teen behavior. A second study, titled "Is the iPhone Birth Control?", focuses on the United States. It highlights a 22% sustained decline in general fertility rates since 2007, unexplained by traditional factors. This research uses the period of the iPhone's exclusive sale on AT&T (2007-2011) as a natural experiment. Findings indicate that iPhone access reduced births among younger women aged 15-24. These cohort effects suggest the iPhone contributed significantly to the decline in births among women under 30. Overall, the diffusion of the iPhone is estimated to explain a substantial portion of the decline in the general fertility rate for women aged 15-44. National survey data on time use and behavior supports the idea that iPhones reduce in-person interactions and sexual frequency.

Polish lawmakers have enacted a new law criminalizing "trash streaming" of serious offenses. Online broadcasts depicting crimes such as rape, murder, animal cruelty, or humiliating violence, including simulated acts, can result in up to five years in prison. This legislation is part of Poland's broader effort to increase online content regulation, following measures like banning child mobile phone use in schools and strengthening pornography age verification. The new law specifically targets the public dissemination of illegal and harmful acts through live streaming platforms.

Coinbase has introduced Coinbase for Agents, a new tool enabling AI agents to conduct cryptocurrency trades and manage payments autonomously. These agents, such as ChatGPT or Claude, can be prompted by users to perform various financial tasks. This includes rebalancing portfolios, identifying trading prospects, executing strategies, and managing positions over time. The platform will eventually extend these capabilities to include trading stocks and making predictions. Coinbase's machine-to-machine payment protocol, named x402, allows agents to directly pay for digital services like research and data APIs without human intervention. This facilitates the execution of trades based on insights gained from these services. The company views this as a step towards agentic shopping, where AI agents will autonomously browse, find deals, and make purchases for users. The core concept behind Coinbase for Agents is to grant AI agents access to financial resources, thereby expanding their capabilities across the internet. Coinbase's AI product lead, Lincoln Murr, highlights the shift towards agents becoming major economic players online, similar to the mobile revolution of the past decade. The x402 protocol, launched in May 2025, has already processed over 100 million transactions. Currently, approximately 157,000 agents are utilizing the protocol for autonomous payments. The significant demand for agents to pay for services independently has underscored their potential as primary financial actors on the internet.

EuroOffice, a new open-source browser-based office suite, has launched its first stable release, aiming to compete with Microsoft 365 and Google Workspace. A consortium of EU companies, including Nextcloud and Ionos, developed EuroOffice to bolster European digital sovereignty through EU-based infrastructure and law. The suite focuses on document editing and requires integration with other platforms for features like storage and sharing. While EuroOffice can be self-hosted on Linux servers, pre-packaged solutions are available from companies like Nextcloud and Ionos. These initial offerings are web-based, not standalone desktop applications. The project aims to provide European organizations with a familiar user experience while maintaining data control within the EU. However, The Document Foundation (TDF), which stewards LibreOffice, criticizes EuroOffice's default use of Microsoft's OOXML format. TDF argues that OOXML's deep ties to Microsoft hinders true open standards and digital sovereignty. They assert that compatibility with OOXML inadvertently supports Microsoft's lock-in strategy.

A man in Florida is suing police and law enforcement agencies after being wrongly accused of attempted child abduction based solely on an AI facial recognition match. Robert Dillon was identified by the software with 93 percent confidence using low-resolution, grainy cell phone photos of a security camera screen. The American Civil Liberties Union (ACLU), representing Dillon, argues that the facial recognition system's accuracy is significantly degraded by poor image quality. The lawsuit highlights that the initial images were secondhand, taken from a screen with a cell phone, and a copy of the original footage was not preserved. Investigators reportedly used these subpar images to query the facial recognition system, leading to the flawed identification. The ACLU emphasizes that facial recognition results should not be the sole basis for an arrest warrant. Even the Jacksonville Sheriff acknowledged that a facial recognition hit alone would not be sufficient probable cause for an arrest. The lawsuit names the city of Jacksonville Beach, individual police officers, and officials from the Jacksonville Sheriff's Office. This case underscores the potential dangers of relying on unverified AI technology in law enforcement investigations. The legal action seeks to hold accountable those responsible for the man's wrongful arrest and ordeal.

OpenAI is reportedly planning significant price reductions for its AI models due to increasing competition from Anthropic. Both AI companies are vying for users as they prepare for potential initial public offerings. The Wall Street Journal indicated that OpenAI is anticipating similar price cuts from Anthropic. OpenAI has a tiered subscription model for its models, while Anthropic offers Claude Pro and Claude Max subscriptions.

Opendoor is closing its India operations after less than two years, according to CEO Kaz Nejatian. Nejatian stated that Opendoor's customers are in America, and that is where its operational work should be located. This decision is seen by some as an early indicator of how AI is changing the economics of global outsourcing. Investors believe AI will replace manual work, leading to job losses in India's outsourcing sector. Venture capitalists describe the move as a watershed moment for AI-driven operations. They argue that AI advancements are challenging the cost advantages that made India a popular offshoring destination. Experts note that AI is reducing the overall need for operational labor, enabling leaner organizations. This trend is not isolated but part of a broader pattern of AI-powered operational redesign. The winners will be companies that integrate AI, software, and human expertise for efficient outcomes. This shift to "Services-as-Software" is expected to impact India's vital export industry.

Xbox's leadership has informed staff that the current financial model is unsustainable. This is due to significant spending exceeding $20 billion over five years without commensurate revenue growth. Excluding the Activision Blizzard King acquisition, annual revenue has paradoxically declined. Executives also highlighted ongoing limitations in console production, impacting their ability to meet player demand. A new business model and hardware partnerships are being considered to address these supply constraints. Furthermore, Xbox faces challenges stemming from an overextended portfolio of acquired studios. The company acknowledges that its growth strategy has led to difficulties in supporting all its content creation teams. In a competitive market saturated with entertainment options, attention is now considered the primary rival. These financial realities suggest substantial workforce reductions are anticipated following the end of Microsoft's fiscal year. The internal memo, while not detailing layoffs, signals a significant shift in Xbox's operational strategy.

OpenAI researchers have identified an online influence operation likely orchestrated by China. This campaign aimed to shape U.S. perceptions of artificial intelligence and influence debates on AI infrastructure. The operation was detected because Chinese-backed operatives used ChatGPT to generate content for social media. Two clusters of ChatGPT users, believed to be from China, created social media posts promoting specific narratives about AI. These narratives included claims that data center construction is increasing electricity costs for Americans. They also promoted the idea that former President Trump used tariffs to gain an advantage in the global tech race. One cluster also used ChatGPT to generate images and comments supporting these claims. These posts were then distributed by accounts impersonating Americans. Another cluster used AI to criticize Trump's tariffs as an attempt to control technological competition. The use of Simplified Chinese prompts and instructions to exclude President Xi Jinping were indicators of Chinese involvement. While the campaigns aimed to amplify existing backlash against AI data centers, they did not achieve significant authentic engagement. The operations are significant for revealing China's intentions and the narratives they are testing.

A report from NewScientist has confirmed that fully autonomous drones with no human oversight have killed soldiers on the battlefield for the first time. This milestone in warfare was achieved during a test involving 10 AI-controlled drones on the front line of the Ukraine war, resulting in the deaths of Russian soldiers. The test was conducted two years ago by a Ukrainian defense industry company, which supplied the technology for the autonomous drones. The drones, known as "Terminator" drones, were programmed to fly towards the front line and engage targets using an AI model. Once launched, the drones operated independently with no connection to human operators, and everything they detected in the target area was killed. The test was carried out by an unnamed military unit near the cities of Bakhmut and Chasiv Yar as part of a Ukrainian counteroffensive push. Human-piloted drones were sent to the area after the test to manually check the results, which included the deaths of a couple of soldiers and the destruction of a truck. The Ukrainian Ministry of Defence has not responded to questions about the test or the current legal position on the use of fully autonomous weapons. The use of fully autonomous drones in warfare marks a significant development, with potential implications for the future of combat. The test has raised concerns about the lack of human oversight and accountability in the use of autonomous weapons, highlighting the need for clear guidelines and regulations on their deployment.

People naturally tend to walk in an anticlockwise direction when ambulating. This tendency was discovered serendipitously by researchers studying social distancing during the pandemic. They observed crowds overwhelmingly moving counterclockwise in videos. Subsequent experiments with individuals and small groups in enclosed spaces confirmed this pervasive observation. The anticlockwise bias was found to be consistent across different cultures, including in Japan. It also held true regardless of handedness, footedness, or eye dominance. Both males and females exhibited this behavior, with a more pronounced bias noted in children. Researchers suggest this bias may stem from subtle biomechanical asymmetries in individuals. The way the brain processes information and coordinates muscles could lead to a slight lean. Right-side dominance might also contribute, particularly in activities like running. While the precise mechanism remains an open question, the anticlockwise walking bias is a persistent finding.

In May 2026, solar power generation surpassed coal for the first time on record in the United States. Solar contributed 12.8% of the nation's electricity, while coal accounted for 12.2%. This represents a significant transformation in the U.S. energy landscape. Just five years prior, coal's share was much higher at 19.7%, with solar at only 5.4%. U.S. solar generation reached a new monthly record of 45.5 terawatt-hours in May 2026, an increase from the previous year. This record is expected to be challenged again this summer due to typical seasonal solar output peaks. Although solar production usually peaks in summer, its share is often highest in spring due to strong sunlight and lower cooling demand. Solar also became the third-largest individual electricity source in the U.S. in May, trailing only natural gas and nuclear power. Conversely, coal generation continues its decline, hitting an all-time monthly low in April 2026. Despite a slight increase in May, coal output remained significantly lower than the previous year, unable to match solar's rapid expansion.

A security researcher known as Nightmare Eclipse has unveiled a new Microsoft Defender zero-day exploit named "RoguePlanet." This exploit reportedly functions on fully updated Windows 10 and 11 systems. It achieves this by creating a command prompt with SYSTEM privileges through a race condition within Defender. The exploit's release occurred shortly after Microsoft's largest Patch Tuesday update to date, which addressed two previously disclosed vulnerabilities. Nightmare Eclipse shared a proof-of-concept for RoguePlanet, having faced repository removals from platforms like GitHub and GitLab. The researcher noted that the race condition exploit's success rate can vary between machines. Cybersecurity firm ThreatLocker confirmed the exploit's viability on patched Windows 11 systems. ThreatLocker suggested that application allowlisting can prevent RoguePlanet's execution. Initially, RoguePlanet was developed as a remote code execution flaw involving Defender's handling of files on remote SMB shares. The researcher explained that successful exploitation previously led to Defender overwriting its own files, resulting in RCE. While another scenario could enable RCE by forcing a victim to open an SMB share, Microsoft reportedly patched a related attack vector in May. Nightmare Eclipse stated that rewriting RoguePlanet to its original functionality was difficult, and it is currently unclear if it can achieve remote code execution.

Visa is partnering with OpenAI to integrate its payment network with ChatGPT. This integration will allow AI agents to make purchases on behalf of users through the ChatGPT interface. Previously, similar initiatives were limited to a small number of merchants. OpenAI's technology will enable AI agents to interact and initiate purchases. Visa will provide the necessary payment authorization and fraud monitoring capabilities for these transactions. The goal is to ensure these AI-driven transactions are trusted, secure, and seamless as AI becomes more active in the economy. An example given involves a customer requesting wireless headphones under $150, which ChatGPT could then find and purchase. Financial terms and specific fees for this collaboration were not disclosed. Visa has stated that safeguards will be in place, including spending limits and approval steps. This aims to protect consumers and minimize fraud during AI-powered shopping.

Valve is discontinuing physical Steam Gift Cards due to persistent scamming issues. While physical cards will no longer be restocked, existing ones can still be redeemed. The company has attempted to address these scams for years through various measures but acknowledges they cannot be entirely eradicated. Steam Digital Gift Cards will remain available and unaffected by this change.

Meta significantly changed its content moderation policies last year, claiming over-enforcement and excessive censorship. The company stated its aim was to allow more legitimate political debate. However, new research from the Center for Countering Digital Hate (CCDH) reveals the consequences of these relaxed rules. Analysis of millions of Facebook comments showed a tripling of abusive and racist comments targeting lawmakers from both parties. Specific categories of harmful content saw even more dramatic increases. Violent threats and hate speech quadrupled in the six months following the policy changes. The report highlighted instances of gendered and racist abuse directed at elected officials that Meta did not remove. Threats against former President Trump more than doubled, with many including direct death threats. Meta's own data indicated a quadrupling of comments violating violent threat policies and hate speech rules. Bullying and harassment comments also doubled after the policy overhaul.

BYD plans to deploy 3,000 ultra-fast "Flash Chargers" throughout Europe by the end of 2027. The initial installations are already underway in Germany and the United Kingdom. This ambitious network is projected to cost approximately $2 billion, with each charger estimated at $670,000. These 1,500kW charging stations are considerably more powerful than Tesla's 500kW V4 Superchargers. However, Tesla already boasts a significant presence with 20,000 chargers across Europe. BYD claims their chargers will minimize grid strain by utilizing batteries that recharge overnight. Any vehicle with a standard CCS charge port can utilize these Flash Chargers. Optimal charging speeds, however, will be exclusive to BYD vehicles featuring the new Blade Battery. Currently, only the Denza Z9 GT, priced at $133,000, is equipped with this battery in Europe. This specific model can achieve a 70 percent charge in just five minutes.

The Asahi Linux team is advising users of Apple Silicon Macs not to upgrade to the macOS 27 beta. This is because recent changes in the beta's boot picker and Startup Disk app render Asahi Linux partitions invisible. Consequently, users will be unable to boot into their Linux installations. The Asahi team recommends installing a secondary copy of macOS 26 or macOS 27 itself on a separate volume before upgrading. Their installer has also been updated to temporarily block installations on macOS 27. They will not support users who install the macOS 27 beta without a stable macOS version present. The Asahi team believes this issue might be an accidental bug rather than a deliberate attempt by Apple to block Linux. They have submitted a bug report to Apple regarding this problem. Fortunately, existing Asahi Linux partitions are not lost; they are simply no longer visible to the macOS boot process. Users who have already upgraded to the beta should not worry about data loss.

A Munich regional court has declared Google directly liable for false claims made in its AI Overviews. The case involved AI Overviews wrongly associating publishers with scams and dubious business practices. The court rejected Google's defense that users could verify information independently. It reasoned that AI Overviews differ from traditional search results by rewriting and judging content in its own words. Therefore, Google creates its own statements, holding it responsible for their accuracy. Existing German legal precedent for traditional search engines, which offered limited liability, does not apply to AI Overviews. This is because AI Overviews generate new, substantive statements by evaluating and combining third-party content. The court emphasized that Google alone has the ability to verify these AI-generated claims by comparing them to the original sources. It also pointed out that AI Overviews are an optional feature, not essential for internet use. Google's argument that users are aware AI-generated information requires caution was also dismissed by the court.

Seattle has implemented a one-year moratorium on new datacenters, becoming the largest U.S. city to take such action amidst growing national concerns about AI infrastructure. The city council unanimously approved the ban, aiming to address the significant electricity demands of these facilities. This pause will allow lawmakers to develop regulations specifically for datacenters, particularly those supporting the AI sector. The moratorium also intends to protect residents from environmental risks and escalating electricity costs. Seattle's mayor indicated that the city will assess if datacenters are an optimal use of urban land. Potential new requirements for datacenters could include investments in local transit and housing. This move is seen as a response to public pressure and a desire by elected officials to address these issues proactively. An amendment permits existing datacenters to seek expansions for up to 20 megawatts of additional power during the moratorium. Critics worry this provision could increase power demand and weaken the moratorium's purpose. Officials defended the amendment, distinguishing between existing civic-purpose datacenters and new AI-focused facilities.

Microsoft's latest Patch Tuesday release is unprecedented, addressing nearly 200 vulnerabilities, far exceeding previous records. This update includes 32 critical flaws and three zero-day vulnerabilities, highlighting a concerning trend. Experts warn that AI is accelerating the discovery of software flaws at an alarming rate. The number of CVEs from Microsoft this year already surpasses the total from 2018. The sheer volume raises questions about the quality of the patches being produced. When considering vulnerabilities in Chrome, Edge, and other third-party software, the total number of fixes approaches 600. This situation has led to discussions about a "Patch Apocalypse" due to the overwhelming number of issues. The accelerated pace of vulnerability discovery is largely attributed to new Large Language Models. Vendors are now releasing patches more frequently and continuously to address these growing threats. This increased speed also means more zero-day and n-day exploits are expected. The time between a patch release and exploitation has dramatically shortened, posing a significant challenge for organizations. Many companies are acknowledging the necessity of AI for their security research and vulnerability remediation efforts.

Commonwealth Fusion has detailed the physics behind its planned 400 MW power plant, ARC, in five peer-reviewed papers. ARC is designed as a tokamak utilizing high-temperature superconducting magnets for fusion reactions. It will fuse deuterium and tritium, producing helium, neutrons, and radiation. The helium acts as "ash," while neutrons and radiation heat a surrounding molten salt blanket. This heat will drive a turbine to generate electricity. The molten salt also contains lithium which, upon absorbing neutrons, produces more helium and tritium fuel. The design aims for approximately 1.13 GW of fusion power, with 500 MW extracted as electricity. After powering itself, ARC is expected to deliver 400 MW to the grid. Fusion will occur in 15-minute pulses, followed by one-minute resets to maintain thermal inertia. This pulsed operation differs from SPARC, which lacks the heat extraction for sustained fusion. The reactor's inner walls will be shielded by tungsten to mitigate erosion. The vacuum vessel is designed for replacement every one to two years, with the tokamak splitting in half for maintenance.

NASA has announced the crew for Artemis III, which will now serve as an Earth-orbit test mission. Randy Bresnik will command the mission, with Luca Parmitano as pilot. Frank Rubio and Andre Douglas will be mission specialists. The crew will test lunar landers developed by SpaceX and Blue Origin. These landers are intended for the Artemis IV mission in 2028. The test flight aims to prove operational capabilities with partner hardware and software. Bresnik has prior ISS experience, including a previous command. Parmitano also has extensive ISS experience and has performed spacewalks. Rubio holds the record for the longest single American spaceflight. Andre Douglas will be making his first spaceflight as an engineer. Bob Hines has been named as a backup crew member for the mission.

The FCC is considering a new rule that would effectively eliminate "burner phones" by requiring telecoms to collect extensive personal information from all customers. This includes government-issued identification numbers and physical addresses for new and renewing customers. The stated goal is to combat scammers and aid law enforcement in identifying criminals. However, privacy advocates and civil rights activists are deeply concerned by these proposals. They argue that this level of data collection mirrors practices in authoritarian countries. Critics contend that it will harm privacy-conscious individuals, domestic abuse survivors, and journalists. The ACLU compares these potential changes to surveillance measures previously seen overseas. They believe this rulemaking could take away the ability to obtain anonymous phone services. Such a move would disproportionately affect vulnerable populations.

The Pentagon has added several Chinese companies, including Alibaba and BYD, to a list of firms it alleges support China's military. This addition to the list, established by Congress in 2021, restricts these companies from receiving U.S. defense contracts. The Pentagon's list targets companies linked to China's defense industrial base, not just those directly controlled by military or security forces. It aims to identify entities contributing to the development of advanced technologies and expertise for China's military. Last year, the Pentagon noted the Chinese military's interest in acquiring technologies from what appear to be civilian Chinese companies and research programs. Both the implicated companies and China's embassy have denied these allegations. The Chinese Embassy criticized the U.S. for an overstretched national security concept and discriminatory practices. They stated that Chinese companies adhere to the laws of the countries where they operate. The embassy urged the U.S. to cease these actions and foster a fair business environment for Chinese firms.

The European Commission has ordered Meta to temporarily restore free access to its WhatsApp Business API for rival AI chatbots. This decision comes as the Commission investigates whether Meta's ban on third-party assistants constitutes an abuse of its dominant market position. Meta intends to appeal this ruling, labeling it as regulatory overreach that would permit major AI firms to use a paid product without cost. The EU initiated its investigation in December following Meta's prohibition of general-purpose AI assistants from the WhatsApp for Business API. They argue this ban appeared to exploit Meta's dominant status in European markets. As an interim measure, Meta has been given five working days to reinstate previous access terms for these third-party AI assistants. The Commission emphasizes the need for interim measures in fast-changing markets to prevent the loss of competition before a final decision is reached. This move aims to preserve user choice regarding AI assistants on WhatsApp. Failure to comply with the interim decision could result in a fine of up to 10% of Meta's total turnover. Meta contends that the Commission's decision allows large global companies to use their paid product for free, subsidized by paying European companies.

Anthropic is launching Claude Fable 5, a powerful AI model for enterprise customers and paid subscribers. This release marks a step towards deploying advanced "Mythos-class" AI models at scale. New safety measures are in place to block high-risk requests in sensitive fields like cybersecurity and biology. Anthropic aims to offer significant benefits while mitigating potential harm through these safeguards. Claude Fable 5 demonstrates exceptional performance, surpassing its predecessor, Claude Opus 4.8, on various benchmarks. The model shows a substantial increase in capabilities, necessitating enhanced guardrails against misuse. If a user poses a dangerous query, Claude Fable 5 will refuse to answer and defer to Claude Opus 4.8 for a safe response. Anthropic has developed new classifiers and safety guardrails specifically for this launch. In parallel, the company has released Claude Mythos 5, which is the same underlying model but with some safeguards removed. This allows for broader application where risks are deemed lower.

A critical Linux kernel vulnerability, CVE-2026-23111, has been discovered in the nf_tables subsystem. This bug allows untrusted users to gain root privileges by exploiting a simple coding error. The vulnerability stems from a misplaced exclamation point within the nf_tables code. This typo creates a use-after-free condition, corrupting memory with malicious code. An unprivileged user can exploit this by manipulating the deletion of "verdicts," which determine actions for network packets. The exploit involves disrupting the deactivation of "catchall elements," acting like wildcards. When a verdict map is deleted, a chain's reference counter should decrement. Errors can reverse this deletion and re-increment the counter. CVE-2026-23111 allows this process to be manipulated, leading to the decrementing of a variable and subsequent deletion of a chain while objects still reference it. Although patched in February, proof-of-concept exploits have already appeared.

The European Commission states that Apple alone decided to delay the launch of Siri AI in the EU. Apple claims regulators rejected its compliant proposals for interoperability. The Commission argues Apple failed to develop solutions meeting EU privacy and security standards. Instead of finding a compliance solution, Apple sought an exemption from Digital Markets Act rules. This exemption request is not considered an option by the Commission. Apple's senior vice president, Craig Federighi, expressed disappointment with regulators' stance. He mentioned their rejection of a "Trusted System Agent" proposal. This system was designed to grant third-party virtual assistants safe access to device capabilities. The Commission disputes Apple's narrative, stating Apple requested a blanket exemption. Apple framed the DMA's requirements as demanding excessive access for third-party AI systems.

Meta is expanding its use of off-platform activity data shared by other businesses to personalize content on Facebook and Instagram feeds, not solely for ads. This change, effective in July, will allow for more tailored content, such as showing camping-related videos if you recently purchased a tent online. Users can disable this feature through the "Activity from other businesses" setting, and Meta clarified that no new data is being collected. Previously, Meta primarily used on-platform activity and AI conversations to personalize ads and content.

Microsoft has recently shut down a significant number of its GitHub repositories, including those associated with Azure and AI coding agents. This action is a direct response to a data breach that has been under investigation. Cybersecurity researchers have identified that hackers injected malware designed to steal user credentials. This malware was embedded within AI coding tools, making it particularly concerning for developers using these platforms. The exact scope of the breach is still being determined, but over 70 Microsoft repositories were disabled. A specific compromised package has been identified as a likely point of entry. These mass repository shutdowns are highly unusual for any company, especially one as large as Microsoft. The affected repositories span various Microsoft divisions, including Azure, their cloud computing service, and AI development projects. Tools like durabletask, used by developers, were also impacted. The breach appears to have originated from a malicious commit to the durabletask repository. This malicious commit introduced configuration files that aimed to harvest credentials when repositories were opened in popular coding environments. Microsoft has confirmed the removals and stated they are investigating potential malicious content. Some repositories have been restored, while others remain offline pending further review. The company has also notified a limited number of customers who may have downloaded content from the affected repositories.

NHS England is planning to roll out Microsoft Copilot to its staff after a successful pilot. The pilot involved 30,000 people and found that the AI assistant saved users an average of 43 minutes a day on administrative work. The rollout will eventually cover 505,000 clinicians and support staff, with each trust receiving a central allocation of licenses based on headcount. The initial allocation will typically be around 2,000 Copilot seats per trust, with more than half a million staff expected to have access by October 2026. The NHS has a significant amount of administrative work that can be handled by the software, including discharge paperwork and data analysis. The rollout also envisions Copilot helping with tasks such as bed management, rota planning, and meeting minutes. In addition to Copilot, NHS organizations will receive access to Copilot Studio, a toolkit for building custom AI agents. This will allow trusts to develop agents for tasks such as handling Freedom of Information requests and processing complaints. A governance framework called Agent 365 will oversee the deployment of these custom systems to ensure their safe and effective use. The rollout is expected to help reduce the administrative workload of NHS staff, allowing them to focus on more critical tasks.

UK Prime Minister Keir Starmer has issued a deadline to tech giants like Apple and Google, demanding they implement device-level safeguards against child exploitation involving explicit images. These protections must prevent children from taking, sharing, or viewing such content. If companies fail to comply within three months, the UK government will introduce legislation mandating these safety features on all new phones and tablets sold in the country. Non-compliant tech firms could face significant fines, and their senior management could be held criminally responsible. Starmer emphasized that this is an achievable challenge and that the government will act if businesses do not voluntarily implement these controls. The Home Office stated that these measures aim to protect children from predators and prevent access to pornography. Adults will still be able to share and view nude content after verifying their age. Parliamentary colleagues have supported this push for regulatory action, arguing that voluntary requests have been insufficient. They believe legislation is the only way to compel social media companies to prioritize child welfare. This initiative is intended to complement the existing Online Safety Act, which already mandates policies for removing illegal and harmful online material concerning children.

Mysterious, brief GPS interference bursts detected across Europe are believed to originate from Russian EKS early-warning satellites. This marks a rare instance of human-made GPS interference originating from space. Researchers identified recurring patterns of high-powered interference lasting less than ten seconds. These signals were simultaneously detected by ground stations across Europe, extending to Greenland and Canada. The interference events occurred predominantly on Tuesdays, Wednesdays, and Thursdays during European business hours. Calculations by the researchers suggest the source must be at least 1,200 kilometers above Earth. Scientists speculate these may be tests of Russian satellite jamming capabilities or communications tests. The signals also interfered with China's BeiDou navigation system. Experts consider this a significant escalation in electronic warfare. While the intent remains unclear, the interference is undoubtedly intentional and could disrupt legitimate GPS services.

An investigation supported by over 20 independent battery experts claims Donut Lab's highly promoted "solid-state" battery is actually a conventional lithium-ion cell. Voltage curves from testing align with high-nickel NCM lithium-ion chemistry, not a sodium-ion solid-state design. Testing also revealed a characteristic "kink" in cell expansion data, indicative of a graphite anode, which sodium ions are too large to inhabit. This strongly suggests the use of lithium ions in the battery. The actual energy density measured was around 298 Wh/kg, falling short of the company's advertised 400 Wh/kg. Donut Lab had raised approximately $25 million from investors based on claims of exceptional performance and rapid charging. The technology appears to originate from CT Coatings, a company with a diverse patent portfolio. CT Coatings had promised a screen-printed sodium-ion solid-state battery to Donut Lab. However, the delivered product was identified as a lithium-ion pouch cell. The evidence from voltage curves and cell expansion data provides compelling proof of its lithium-ion nature.

The world's oceans are facing severe and accelerating pressure from human activities, according to a report from the United Nations. The report highlights the cumulative impact of stressors such as pollution and large-scale industrial fishing, resulting in widespread biodiversity loss and putting ocean systems under severe strain. The UN's third World Ocean Assessment reflects the work of nearly 600 scientists from 86 countries and looks at the oceans' health from 2021-25. The previous report found persistent degradation of the marine environment, and the latest report shows that the damage has worsened in the past few years. The scientists' key findings include a significant increase in sea levels, with the rate of rise doubling over the past decade. The report also notes that 16% of the increase in global ocean heat since 1955 occurred after 2018, with the greatest relative warming observed in the Atlantic Ocean and the southern parts of the Indian and Pacific Oceans. Despite the findings, large gaps in knowledge persist, with only 27% of the ocean floor mapped by 2025, leaving deep-sea ecosystems poorly understood. The report's findings have prompted calls for action, with Greenpeace's global ocean campaigner Lukas Meus urging governments to create fully protected ocean sanctuaries. Governments have promised to protect 30% of the world's ocean by 2030, which is the minimum scientists say is needed for the ocean to recover. The report's assessment is a damning indictment of the impact of human activities on the world's oceans, and it highlights the need for urgent action to mitigate the damage and protect the marine environment.

OpenAI has confidentially filed for an initial public offering, potentially marking a highly anticipated market debut. This move follows similar IPO announcements from competitors like Anthropic and SpaceX. While the timing of the IPO remains undecided, the confidential filing means details on share quantity and pricing are not yet public. OpenAI stated that certain internal objectives might be better pursued as a private entity, but the filing offers flexibility for a public offering if deemed advantageous. Becoming a public company will allow investors to scrutinize OpenAI's financial performance as it invests heavily in AI infrastructure. The company has faced pressure to justify its substantial valuation through revenue generation. Last valued at $852 billion after significant funding in March, OpenAI's financial transparency will increase. This IPO filing comes amid investor concerns about the recent surge in tech stock valuations. The company's ability to demonstrate profitability alongside its ambitious AI development is crucial. The confidential nature of the filing suggests a strategic approach to its public market entry.

Meta had embedded an unreleased facial recognition system called NameTag into its Meta AI app, which is installed on millions of phones. This system was designed to create biometric signatures from faces captured by smart glasses and compare them to a database. Faces the system failed to recognize were cropped, indexed, and stored locally for future processing. Reports indicated Meta was considering a launch this year, despite public statements suggesting no final decision had been made. Following Wired's report, Meta's CTO dismissed the findings as misleading and dishonest. In response to the report, Meta has now removed nearly all traces of the NameTag facial recognition code from the latest version of the Meta AI app. The removed components include the face-recognition software and the alert for identified individuals. A folder for storing unrecognized faces and their biometric signatures has also been eliminated. However, a few fragments of the NameTag system, such as a debug menu label and a dormant link, still remain. These remnants suggest the system is no longer functional within the app.