Welcome to the BrevFeed daily tech briefing for Wednesday, September 23. We've got 21 stories for you today across A.I., security, software, cloud, hardware and startups. Let's get into it.

In AI news, new research from Boston College indicates that older workers are leaving jobs in AI-exposed industries at a higher rate. This trend suggests that AI's impact on employment isn't limited to younger generations, but also significantly affects those aged 55 and above.

The study, led by Professor Geoffrey Sanzenbacher, found that these older workers are either facing unemployment or voluntarily changing jobs due to AI implementation. While AI can automate tasks and potentially displace some roles, it also offers opportunities for augmented work and increased productivity, which could even extend some careers.

This research adds a new dimension to the ongoing discussion about AI's influence on the job market. It highlights how automation pressures might lead some older employees to seek non-AI roles or consider early retirement, while others might benefit from AI-driven productivity gains.

Turning to security, the hacking group ShinyHunters claims to have breached FBI-related services, obtaining personal data for all FBI employees and applicants. This includes sensitive information like names, home addresses, and phone numbers. The group provided a sample of 5,000 employee records, which 404 Media partially verified, finding some phone numbers linked to U.S. Department of Justice personnel.

ShinyHunters also defaced the FBI jobs website, which is currently unavailable. The group claims to have stolen between two and three terabytes of data by exploiting a new Oracle PeopleSoft zero-day vulnerability, which allowed them to access an Amazon-hosted government cloud. They say they compromised FBI Criminal Justice, HR, and Medlink services.

The hackers state their motivation is not financial, but rather a demand for the FBI to retract a May 2026 public service announcement that detailed ShinyHunters' targeting of Canvas. They are giving the FBI one week to remove the report. This alleged breach carries significant national security and counterintelligence implications, as such data could be used by criminals or foreign intelligence agencies.

In AI, Alphabet reported strong second-quarter revenue of nearly 120 billion dollars, a 24% increase year-over-year. This marks their twelfth consecutive quarter of double-digit growth, with earnings per share beating Wall Street expectations.

Google Cloud was a major driver, with revenue up 82% to 24.8 billion dollars, largely due to enterprise AI solutions and infrastructure adoption. The company's cloud contracting backlog also climbed to 514 billion dollars.

However, Alphabet reported negative free cash flow of 5.9 billion dollars, its first in over a decade, due to massive AI infrastructure investments. They've increased their yearly capital expenditure forecast to between 195 and 205 billion dollars, with quarterly capital spending up 100% year-over-year, mostly on servers and data centers.

Following the report, Alphabet shares fell 3 to 4 percent, as investors reacted to the mounting costs of AI development. CEO Sundar Pichai emphasized that these investments are crucial for redefining possibilities across all business areas, highlighting the ongoing race in AI among major tech companies.

Turning to security, the FBI's New Orleans field office has launched an official inquiry into a new dark web service called Nexus. This service is reportedly selling digital scans of over 153 million driver's licenses from individuals in the United States and Canada. The data also includes other identification cards and medical records.

The images appear to originate from a Louisiana-based identity verification company, which has now confirmed a data breach. This company provides identity verification solutions to a wide range of businesses, including Fortune 500 companies, car rental agencies, and financial institutions. The breach highlights a significant vulnerability in identity verification processes, posing a substantial risk for identity theft for millions of individuals.

The dark web service, Nexus, claims to have been exfiltrating data from the identity verification company for over a year, adding approximately half a million new documents daily. The Privacy Commissioner of Canada has also launched an investigation into the company's security practices and victim notifications, assessing its compliance with federal privacy laws.

In AI news, Anthropic has launched a new Reflect dashboard for its Claude chatbot, giving users insights into their AI interactions. This tool provides analytics on topics discussed, peak usage times, and tasks handled by the AI, allowing users to track their usage over various timeframes, from one month to a year.

Inspired by features like Spotify Wrapped, the Reflect dashboard aims to help users consider if their engagement aligns with personal goals. It also promotes mindful use by allowing users to set 'quiet hours' and break reminders, encouraging reflection on how often and why they use the AI.

Additionally, 1Password has introduced a feature allowing Claude to access user credentials without exposing them. This 'zero-exposure security framework' ensures passwords and two-factor authentication codes do not interact with Anthropic’s AI systems, enhancing Claude’s ability to automate tasks securely and efficiently without compromising user privacy.

In hardware, Nvidia's RTX 50-series graphics cards are seeing significant price increases globally, with some models now costing 10% to over 50% above their initial MSRP. This trend is evident across China, South Korea, Japan, and the US, impacting both high-end and more budget-friendly GPUs.

These price hikes are largely attributed to rising component costs, including TSMC wafers and GDDR7 memory. For example, market research firm TrendForce notes that GDDR7 2GB modules, crucial for the RTX 50 series, have increased to 20 dollars per unit. Nvidia has reportedly passed on higher GPU package costs to its board partners, who then transfer these increases to consumers.

Another major factor is the high demand for AI hardware, which diverts component supply and drives up manufacturing costs for consumer GPUs. This has led to situations where even Nvidia's workstation GPU, the RTX PRO 6000 Blackwell, has seen its price double to 16,000 dollars. The impact is so widespread that even older cards like the RTX 3060 12GB, which Nvidia re-released, have seen price jumps of up to 45% in just two months.

In security, new research from the Hong Kong University of Science and Technology reveals significant vulnerabilities in AI coding agents like OpenAI Codex and Claude Code. They've developed a technique called SKILLCLOAK, which can bypass AI skill scanners over 90% of the time. This allows malicious code to evade detection, even though these scanners are designed to identify and block potential threats in AI agent marketplaces.

The researchers also found that when these AI coding agents operate in autonomous mode, they can be tricked into executing malicious code instead of flagging or blocking it. This means harmful commands could run on a user's machine. There's currently no patch for this, so developers are advised to use non-autonomous settings until a solution is found.

Further studies highlight other attack vectors, including agent data injection, which corrupts an AI agent's input data, leading to unauthorized actions. Additionally, vulnerabilities have been found in open-source mobile AI agent frameworks, where malicious attacks can be carried out using invisible screen text. These findings underscore the urgent need for more robust security measures and rigorous testing in AI-driven technologies.

In AI policy, the U.S. has proposed a new notification mechanism to China for artificial intelligence incidents that could affect national security. This initiative aims to increase transparency between the two leading AI powers, particularly concerning cross-border AI activities.

The proposal was made during recent discussions between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng, ahead of upcoming talks between President Trump and China's Xi Jinping. Bessent emphasized the need to move from opacity to greater transparency regarding AI threats and goals.

This notification system is designed to prevent misinterpretations of autonomous AI actions and address the risk of rogue AI models. The U.S. and China are also setting up a new "Board of Trade" and discussing potential tariff reductions on various goods, indicating a broader diplomatic effort.

In cloud infrastructure, Kubernetes is now being considered for managing desktop infrastructure, a significant shift from its traditional role with containerized applications. This move aims to unify operational practices and potentially lower costs associated with outdated virtual desktop systems.

Historically, desktop infrastructure has been managed separately, often relying on pre-allocated virtual machine pools. Integrating desktops into a Kubernetes-managed environment could bring them into a more modern, cloud-native architecture.

However, while Kubernetes simplifies application deployment, it also exposes complexities in managing databases. This requires expertise beyond standard DevOps skills, posing a challenge for teams that may need to acquire new database management capabilities or specialized solutions.

In cybersecurity, Claude Mythos 5, an AI model designed for security applications, is now available within Claude Security for enterprise customers. This integration allows users to scan codebases for vulnerabilities and receive suggestions for patches. The model is also being integrated into various partner cybersecurity tools, making its advanced capabilities more widely accessible through existing defensive platforms.

This expansion follows Project Glasswing, which initially provided early access to Mythos 5. The company has developed safety classifiers and safeguards to ensure that while the AI offers defensive outputs, direct access that could lead to malicious use is restricted. This approach aims to broaden access to advanced AI for defensive cybersecurity while maintaining necessary safeguards.

Additionally, a new initiative called the Defender Advantage Fund has been launched with 35 million dollars in credits. This fund will support organizations focused on patching vulnerabilities in open-source projects, automating scanning and patching processes, and exploring new security approaches to strengthen the security of widely used open-source components. The company also plans to expand its Cyber Verification Program, which offers vetted defenders reduced safeguards on its Opus model.

In international tech policy, the Trump administration is preparing to confront Australia over its proposed digital duty of care laws. These new regulations would fine digital platforms over 100 million dollars for failing to remove harmful content and for not allowing users to opt out of algorithm-generated feeds.

The proposed Australian laws aim to protect children from content like eating disorders, misogyny, and cyberbullying, and would require platforms to offer users a choice between personalized feeds and those showing only direct connections. The US administration views such measures as potential extortion on American tech companies.

Australia's communications minister, Anika Wells, has downplayed the US statement, asserting Australia's right to defend its citizens. The US embassy in Canberra has already submitted concerns, suggesting the laws could reduce the reach of independent journalists or controversial topics. This situation highlights growing international tension over regulating major tech companies.

In cloud computing, SpaceX has filed an FCC application for an orbital data center constellation of up to one million satellites. The project, named Starmind, aims to make space a cost-effective location for AI processing within two to three years, according to Elon Musk.

SpaceX plans to use NVIDIA's Rubin GPUs and Vera CPUs for these Starmind AI1 satellites, adapting the Vera Rubin NVL72 rack-scale AI platform for orbital deployment. Initial launches are targeted for the fourth quarter of 2027, with the project likely depending on Terafab, a joint chip manufacturing plant with Tesla.

However, the ambitious plan faces significant technical challenges, including satellite launch logistics, heat dissipation, and radiation management in space. Doubts also remain about the feasibility of reaching such a large scale and the aggressive timeline, given current manufacturing and launch capacities.

In AI, AMD is making a significant move, announcing a strategic partnership with Anthropic that includes an investment of up to five billion dollars. This collaboration aims to bolster Anthropic's computing infrastructure, with a focus on deploying AMD's Instinct MI450 Series GPUs.

The plan involves installing up to two gigawatts of these AMD GPUs within Anthropic's systems, with the first gigawatt expected to be deployed in the first half of 2024. This partnership also includes a multi-year engineering collaboration, where AMD will integrate Anthropic's Claude AI model into its software development.

This deal highlights AMD's push to compete with other major players in the AI hardware market, like Nvidia. For Anthropic, it means a substantial boost to its AI capabilities and strengthens its position in the rapidly expanding AI infrastructure landscape.

In cloud news, AWS Elastic Beanstalk has introduced a new Cluster Mode, allowing multiple applications to share infrastructure managed by Amazon EKS.

This update aims to reduce per-application costs and operational complexity, especially for teams managing a portfolio of applications. AWS continues to handle the operational responsibility, including deployment, scaling, and monitoring.

The fully-managed Cluster Mode continuously deploys, scales, and upgrades applications, providing a unified experience with consistent operational guarantees, while leveraging the benefits of shared resources.

In AI, a new open-source model called Laya has been released, offering a faster and more cost-effective alternative to TypeSafe AI's proprietary Jev model. Laya is a non-autoregressive AI designed for rapid, schema-based decision-making.

Laya boasts impressive performance, achieving inference times of just 32.8 milliseconds on a single GPU, which its developers claim is six to eight times faster than Jev. It also supports over 100 languages and is released with open-source Apache 2.0 weights, meaning no API costs.

This new model aims to address the limitations of using large generative AI for simple, reflex decisions, which can be a bottleneck in AI pipelines. Laya's open-source nature and speed present a compelling option for developers seeking efficient and transparent AI decision systems.

Turning to security, CISA has issued an urgent alert regarding multiple actively exploited vulnerabilities in Microsoft SharePoint Server. These flaws, including CVE-2026-45659, CVE-2026-50522, and CVE-2026-58644, are primarily related to the deserialization of untrusted data.

The critical issue is that these vulnerabilities allow attackers, often with minimal permissions, to execute arbitrary code remotely on unpatched servers. For instance, CVE-2026-45659 can be exploited by an authenticated attacker with just site member permissions, while CVE-2026-50522, with a CVSS score of 9.8, can lead to unauthorized access and data compromise.

Microsoft released patches for these issues in its July 2026 security updates, covering various SharePoint versions. CISA has mandated that federal agencies apply these patches by specific deadlines under its Known Exploited Vulnerabilities catalog. All organizations using SharePoint are strongly advised to implement these updates immediately to protect their systems from potential breaches.

In hardware, Qualcomm has unveiled its latest flagship smartphone processors, the Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6. Both chips are built on a 2-nanometer manufacturing process and are designed to power the next generation of premium Android smartphones.

A primary focus for these new Snapdragon chips is enhanced on-device AI capabilities. They include new sensing hubs that can run small AI models locally, enabling functions like personal scribes that differentiate speakers and build memory for task automation. The Extreme version can even run a 30-billion-parameter AI model directly on the device, acting as an AI hub without relying on cloud services.

Beyond AI, the chips feature an eight-core Oryon CPU, with the standard Elite Gen 6 offering a 10% CPU performance improvement and 37% better power efficiency over its predecessor. The Extreme variant sees a 13% CPU performance gain. An re-architected Adreno GPU provides up to 35% better GPU performance and 40% improved power efficiency, with the Extreme version also boosting video capture capabilities and gaming performance.

In hardware, PC component prices are surging, reversing decades of declines. The cost of RAM, SSDs, and hard drives has increased significantly over the past year, primarily due to high demand from AI infrastructure buildouts. For example, some 32-gigabyte DDR5 memory kits have jumped from around 90 dollars to over 450 dollars in just one year.

This trend is a historical anomaly, as memory prices have consistently fallen over time. Some RAM prices are now back to levels last seen in 2007 or 2008. The demand for High Bandwidth Memory, or HBM, used in AI training, is a major factor, as it shares manufacturing components with standard DRAM.

The rising costs are impacting PC builders and leading some manufacturers, like Micron, to shift focus from consumer memory to enterprise AI solutions. While the overall outlook suggests continued price increases for the rest of the year, consumers can still find deals through combo offers, price tracking, and refurbished outlets.

In hardware, Apple has updated its Mac Mini lineup, introducing models with the new M6 and M5 Pro chips. This marks a significant performance upgrade from the previous M4 and M4 Pro versions. However, the base M6 model now starts at 899 dollars, a 300-dollar increase from the M4 model's original 599-dollar price point.

The external design of the Mac Mini remains unchanged from its 2024 redesign, maintaining its compact form factor. Connectivity has seen improvements, with the addition of Wi-Fi 7, Bluetooth 6, and 2.5 Gigabit Ethernet as standard features. The M6 configuration also includes two USB-C ports on the front and three Thunderbolt 4 ports on the back.

While the M6 chip offers generational performance improvements, the higher price has led some to question the Mac Mini's previous value proposition. The M4 Mac Mini was known for its affordability, but the new pricing, potentially due to rising component costs, shifts its position in the market. Additionally, unlike the M4, the new M6 and M5 Pro models feature soldered SSDs, similar to earlier M1 and M2 versions, meaning no aftermarket storage upgrades are possible.

In media news, Paramount has settled an antitrust lawsuit with California and 11 other states, clearing a major hurdle for its proposed 110 billion dollar merger with Warner Bros. Discovery.

The agreement includes commitments for Paramount to release at least 30 theatrical films annually for the next five years, with a portion being independent and blockbuster titles. They will also increase U.S. production spending by at least 300 million dollars.

While the settlement paves the way for the merger, which is expected to close in early October, it has drawn criticism from some Democrats and media advocates who question the effectiveness of such behavioral remedies in large mergers.

In automotive tech, a recent comprehensive test has identified the best free CarPlay apps for music and navigation. Google Maps remains a top choice for navigation, while Apple Maps has significantly improved, offering more engaging voice directions and Apple Watch integration. CarPlay itself has evolved into a highly desired feature, projecting smartphone functions onto car infotainment screens for easier use while driving.

CarPlay offers both wired and wireless connections, with accessories available to enable wireless functionality in older vehicles. While CarPlay is available in over 800 vehicle models, some manufacturers like Tesla and Rivian have never supported it. General Motors is even removing CarPlay and Android Auto from its 2025 Chevrolet electric vehicles, opting instead for Android Automotive OS, which is built directly into the car's hardware.

Looking ahead, CarPlay is set to receive new features with iOS 27 in September, including a dedicated Siri app and Siri AI. Meta AI for iPhone is also in limited testing for CarPlay, joining other third-party chatbots like ChatGPT. For those without built-in CarPlay, portable screens offer an alternative, allowing access to iPhone apps and bypassing proprietary automaker systems, though they don't match the quality of integrated systems.

That's the BrevFeed daily briefing. We'll be back tomorrow with the stories that matter in tech. Thanks for listening.