Wednesday, July 15, 2026

Microsoft patches record number of security vulnerabilities, citing its use of AI

Microsoft has released a record 570 security patches this week as part of its monthly 'Patch Tuesday' update, leveraging AI to enhance vulnerability discovery. Among these patches, two are classified as zero-days, meaning they were exploited prior to Microsoft's awareness. One vulnerability allows hackers to escalate privileges on Windows Server, while another affects SharePoint, prompting warnings from the U.S. Cybersecurity and Infrastructure Security Agency (CISA) about active exploitation. This surge in patches follows Microsoft's recent announcement that it anticipates a higher volume of security updates due to AI advancements in identifying dormant vulnerabilities.

OpenAI finally launches hardware… for Codex

OpenAI has officially launched the Codex Micro, a hardware device designed to enhance its coding platform, Codex. This square-shaped device, developed in collaboration with keyboard maker Work Louder, features 13 mechanical switches, a joystick, and a dial, allowing users to manage their Codex agents more effectively. Priced at $230, the Codex Micro includes six frosted keys that provide real-time status updates on tasks, such as completion and errors. This release is distinct from OpenAI's ongoing hardware project with Jony Ive, which is speculated to be a smart speaker expected to launch next year amidst a lawsuit from Apple over hardware secrets.

America made 1,200 new millionaires every day last year — the 4 disciplined habits legendary investors say actually work

In 2025, the United States added around 440,000 new millionaires, translating to about 1,200 daily, according to the UBS Global Wealth Report. This surge in wealth is attributed not to luck or sudden stock picks, but to consistent long-term investing strategies, including contributions to retirement accounts and diversified portfolios. The report highlights that 79% of gross household wealth was linked to rising markets, indicating that wealth accumulation is often a gradual process rather than an overnight success.

Apple Intelligence approved for launch in China with Alibaba’s Qwen AI

Apple Intelligence, the company's generative AI service, has received approval for launch in China, facilitated by a partnership with Alibaba to integrate its Qwen AI model into Apple's operating systems, including iOS and macOS. This development follows a significant 28% increase in Apple's sales in Greater China, totaling $20.5 billion in the second quarter of 2026. Apple also regained its position as the No. 2 smartphone vendor in China after a recent shopping festival boosted iPhone sales. Prior to this collaboration, Apple faced challenges in adapting its AI models for the Chinese market.

Todd Blanche says he has 'full faith' in Kash Patel as FBI director

During his Senate Judiciary Committee confirmation hearing, Todd Blanche, acting attorney general since April 2026, defended FBI Director Kash Patel amid sharp questioning from Senator Sheldon Whitehouse. Whitehouse criticized Patel's travel activities, questioning their appropriateness and suggesting potential misconduct. Blanche responded affirmatively, stating, 'I have full faith in Director Patel and the work that he's doing every day.' The hearing also addressed Blanche's handling of Epstein-related documents and his cryptocurrency holdings, which have drawn scrutiny from Democrats.

Bonsai 27B is a full open reasoning model that fits on an iPhone

PrismML has unveiled Bonsai 27B, a 27-billion-parameter AI model capable of running directly on iPhones without compromising its reasoning and agent capabilities. This model, based on Alibaba's Qwen3.6-27B, supports multistep reasoning, tool use, and image understanding. Notably, the smaller variant of Bonsai 27B is only 3.9 GB, allowing it to fit within the storage constraints of devices like the iPhone 17 Pro Max. PrismML's innovative compression technology enables this efficiency while retaining up to 95% of the original model's performance, making it suitable for local AI applications.

Bank of Canada holds rates, says economy is growing again

The Bank of Canada announced on July 15, 2026, that it would keep its benchmark overnight rate unchanged at 2.25%, marking the sixth consecutive meeting without a rate change. The central bank expressed optimism about economic growth, predicting a stronger second half of the year as inflation pressures ease. However, it lowered its growth forecast for 2026 to 0.7%, down from 1.2%, reflecting a weaker start to the year due to disruptions from the Middle East conflict and U.S. trade uncertainties. The bank also raised its inflation forecast for 2026 to 2.5%, indicating a cautious approach to rising oil prices.

We're buying more of Jim Cramer's favorite chip stock for the second time this week

Jim Cramer's Charitable Trust is purchasing an additional 100 shares of Intel at approximately $110 each, raising its total holdings to 1,100 shares. This move increases Intel's weight in the portfolio from 2.7% to about 3%. The decision follows a positive market reaction to ASML Holding's strong earnings, which highlighted Intel Foundry's use of advanced lithography machines for its latest chip processing technology. Additionally, Intel announced a significant investment of $5.7 billion in its Ireland facility to enhance production capabilities for its Xeon server processors.

The US is advancing AI safety through state and federal action

The US is advancing AI safety through a collaborative effort between state and federal governments, with states like California, New York, and Illinois leading the charge in establishing safety legislation. This initiative, termed 'reverse federalism,' aims to create a unified national standard for AI governance, as the federal government develops a framework for testing advanced AI models. OpenAI is engaged in discussions with the Trump Administration to finalize this framework by early August 2026, ensuring that safety measures are in place to protect against potential misuse of AI technologies.

Anthropic, Blackstone bet the next trillion-dollar AI business is implementation, not models

Ode, a new AI implementation company launched by Anthropic in partnership with Blackstone, is positioned to capitalize on the growing need for enterprise AI adoption, boasting a valuation of $1.5 billion. This venture, which employs 100 engineers, focuses on creating tailored AI solutions for businesses, leveraging technology from Anthropic while also remaining open to other AI products. Ode was built upon the acquisition of Fractional AI, a startup recognized for its effective AI engineering services, highlighting the importance of quality implementation in the rapidly evolving AI landscape.

Watch Fed Chairman Kevin Warsh testify live before Senate banking committee

Federal Reserve Chairman Kevin Warsh is testifying before the Senate Banking Committee, addressing questions about the economy and interest rates. This appearance is part of his mandated congressional obligations. During his earlier testimony to the House Financial Services Committee, Warsh reiterated the Fed's commitment to combating inflation but offered limited insights into future monetary policy. Despite attempts by legislators to draw him into discussions on fiscal and political matters, Warsh maintained a focus on the Fed's core responsibilities.

ASML raises outlook on strong orders for advanced chipmaking equipment

ASML has raised its full-year sales outlook following strong second quarter results, driven by heightened demand for advanced chipmaking equipment amid the AI semiconductor boom. The company now anticipates 2026 net sales of approximately $49.2 billion to $51.5 billion, significantly above the $45 billion average analyst estimate. CEO Christophe Fouquet highlighted that customers are increasing capital expenditures and securing long-term agreements, leading to a backlog of orders for ASML's EUV lithography machines extending into 2028. This surge in demand reflects robust investment in AI infrastructure by leading chipmakers.

Alibaba’s U.S.-listed shares rise 4% after Qwen AI set to be integrated in Apple Intelligence

Alibaba's U.S.-listed shares rose 4% in premarket trading after the company announced that its Qwen AI model will be integrated into Apple services in China. An Alibaba spokesperson confirmed that Qwen will enhance Apple Intelligence experiences across iOS, iPadOS, macOS, and visionOS. This integration follows the Cyberspace Administration of China's approval of Apple AI services, which also includes offerings from local companies like Huawei. The deal comes amid escalating technological competition between the U.S. and China, particularly in the AI sector.

Stripe Benchmark Shows AI Agents Build Integrations but Struggle with Validation

Stripe has launched a benchmark suite to assess AI agents' capabilities in developing full Stripe integrations, focusing on end-to-end workflows that require execution, testing, and validation. The benchmark includes 11 environments that simulate real-world integration projects, with agents evaluated on tasks ranging from backend-only operations to full-stack workflows. Results show that while AI agents like Claude Opus 4.5 achieve a 92% success rate on certain tasks, they struggle with validation and cross-system interactions, highlighting a critical gap in their ability to manage complex integration scenarios effectively.

Report: Apple looking into buying chip startups to strengthen its AI infrastructure

Apple is actively pursuing acquisitions of semiconductor startups to bolster its AI server capabilities, as reported by The Information. The tech giant has engaged with bankers and potential targets to assess their willingness to sell. This initiative arises from challenges Apple faces with its internal AI servers, which currently rely on M2 Ultra chips. The anticipated launch of Apple's next-generation server chip, code-named Baltra, has been postponed beyond its original 2026 timeline, prompting the company to seek external solutions to enhance its AI processing capabilities.

Samsung Health won’t delete all of your data if you opt out of AI training after all [U]

Samsung Health has updated its policy regarding user data in relation to AI training. Users can now opt out of having their health data used for AI development without losing all their information. In a statement, Samsung clarified that only the data collected for AI purposes will be deleted, allowing users to retain their existing health data for continued use of the app. This change comes as part of a broader redesign of Samsung Health, which introduces more AI-powered features, just ahead of the Galaxy Watch 9 launch.

Report: How an email mistake derailed talks between Apple and OpenAI ahead of the lawsuit

Negotiations between Apple and OpenAI regarding alleged trade secret theft deteriorated after an email error by Apple's outside counsel, Gabriel Gross. On July 15, 2026, Apple filed a lawsuit against OpenAI and two former employees, Chang Liu and Tang Tan, accusing them of stealing confidential hardware information. Apple contends that it initially reached out to OpenAI in February to address concerns about its confidential data being misused, but OpenAI failed to respond. OpenAI disputes this, claiming it did reply, but communication broke down when Gross mistakenly emailed the wrong OpenAI employee, leading to confusion and accusations of dishonesty.

Presentation: Postgres for Production Agents: Your Relational Foundation for Enterprise AI

At the QCon AI conference, Gwen Shapira detailed how PostgreSQL is being utilized to scale AI features in mission-critical applications. She emphasized the database's multi-modal capabilities, such as JSONB parsing and high-recall HNSW vector indexing, which enhance the deterministic and semantic context provided to large language models (LLMs). Shapira also discussed vector quantization techniques that can accelerate query performance by up to 4x, alongside strategies for managing agentic memory in AI systems.

OpenAI’s hardware device may partly compete with AirPods Ultra – and lose

OpenAI is reportedly working on a mobile, screen-free smart speaker that incorporates a camera and advanced AI models, which could position it as a competitor to Apple's upcoming AirPods Ultra. While the device aims to deliver a new kind of technology, it resembles existing smart speakers, raising questions about its uniqueness. In contrast, AirPods Ultra will feature cameras in both earbuds to enhance Siri's contextual capabilities, potentially making them a more appealing choice for users seeking convenience and portability in their audio devices.

OpenAI's Codex now encrypts instructions between AI agents, leaving developers blind to internal delegation

OpenAI's Codex has implemented mandatory encryption for the internal instructions exchanged between AI agents, leaving developers unable to track task delegation. Since early June, the main agent's communication to subagents now appears as unreadable strings in session history, a significant shift for developers who rely on visibility to manage tasks. This change affects the larger GPT-5.6 models, Sol and Terra, while the smaller Luna variant still permits readable task descriptions. Developers have reported that this encryption can lead to failed handoffs, particularly when both agents use the same model, raising concerns about the reliability of the system.

Starlink’s V5 dish is now available — here’s how it compares

SpaceX has introduced the Starlink V5 residential dish, which is now available in select areas. This new dish is significantly smaller and lighter than its predecessor, the V4, and boasts improved energy efficiency. The V5 dish measures 384 mm x 306 mm x 34 mm and weighs just 1.1 kg, compared to the V4's dimensions of 594 mm x 383 mm x 39.7 mm and weight of 2.9 kg. While the V5 offers peak download speeds of 375+ Mbps, slightly lower than the V4's 400+ Mbps, it consumes only 35 to 50 watts of power, a notable reduction from the V4's 75 to 100 watts. SpaceX plans to increase production to expand the dish's availability globally.

Meta employees sue over layoffs they say were driven by discriminatory AI selection systems

Former and current Meta employees have filed a lawsuit in federal court in California, alleging that the company used internal AI systems to create layoff lists that discriminated against certain groups. The lawsuit claims that during a May 2026 layoff of 8,000 employees, the AI systems disproportionately affected individuals with disabilities and those on medical or parental leave. One plaintiff reported receiving layoff notification just two days before giving birth. Meta has denied these allegations, asserting that all personnel decisions are made by human managers. The plaintiffs are seeking a preliminary injunction to keep their jobs while the case is arbitrated.

OpenAI Staffers Are Funding a Rival Super PAC to Take on Their Boss

A coalition of OpenAI employees has contributed more than $215,000 to Guardrails Alliance, a super PAC advocating for stricter regulations on AI development. This initiative, which launched last month with $5 million in initial funding, aims to counter the influence of Leading the Future, a pro-AI industry super PAC backed by over $100 million from tech leaders, including OpenAI's president Greg Brockman. The donations reflect rising tensions within OpenAI regarding the company's political affiliations and the push for responsible AI governance.

Nokia’s AI-RAN platform: a radio comeback that runs on NVIDIA

Nokia unveiled its AI-RAN platform on July 15, 2026, claiming it to be the industry's first GPU-accelerated AI-native radio access network. Built on its anyRAN software and NVIDIA's Aerial system, the platform is designed to enhance spectral efficiency, with a goal of achieving over 100% gains by 2028. This launch is pivotal for Nokia, as the mobile business has struggled under CEO Justin Hotard, who has consolidated it into a new Mobile Infrastructure segment. The platform's commercial availability is projected for 2027, with pilot programs set to begin by the end of this year.