Section

Technology

AI, platforms, startups, internet

TechCrunch

OpenAI's Astra Model Introduces New Reasoning Technique Raising Concerns Among AI Safety Experts

OpenAI's Astra model will implement a reasoning technique called 'recurrent depth,' which may complicate the monitoring of its reasoning process, according to a report by The Information. AI safety experts, including Redwood CEO Buck Shlegeris and advocate Zvi Mowshowitz, have expressed concerns about the implications of this technique for AI safety and the potential need for regulations. OpenAI maintains that the use of this technique is limited and emphasizes its commitment to maintaining legible chains of thought.

Bias: 16 Sentiment: -0.20
The Verge

Google launches Gemini 3.8 Flash model with enhanced capabilities

Google has introduced the Gemini 3.8 Flash model, which enhances reasoning capabilities compared to its predecessor. The pricing remains unchanged, but the new model may result in higher costs due to increased token usage for performance optimization.

Bias: 30 Sentiment: +0.00
Ars Technica

NASA Modifies Lunar Spacesuit Design for Artemis IV Missions

NASA has announced a change in the design of its lunar spacesuit, opting for a simpler version called the 'Sortie Suit' for its upcoming missions to the Moon's South Pole, planned for as early as 2028. This decision aims to reduce the suit's mass and complexity while improving compatibility with lunar landers from SpaceX and Blue Origin.

Bias: 4 Sentiment: +0.00
The Verge

1Password faces customer backlash over funding Linux project

1Password has received criticism from customers after pledging $300,000 to support a Linux distribution associated with David Heinemeier Hansson, who has made controversial statements regarding ethnic minorities. The backlash has prompted discussions on social media about alternative password management options.

Bias: 4 Sentiment: -0.50
Wired

U.S. Government Supports OpenAI in Copyright Dispute with New York Times

The U.S. government has sided with OpenAI in its copyright dispute with the New York Times, arguing that the use of copyrighted materials for AI training may fall under fair use. The letter submitted to the court claims that the Times' interpretation of fair use is inconsistent with current law and warns of negative consequences for AI development. This case is part of a broader legal landscape involving multiple lawsuits concerning AI and copyright issues.

Bias: 4 Sentiment: +0.00
Hacker News — Front Page

Comparison of Async Rust and FreeRTOS on STM32F446 Microcontroller

This article compares async Rust using Embassy with FreeRTOS using C on an STM32F446 microcontroller, focusing on performance metrics such as interrupt latency, program size, and RAM usage. The results indicate that Embassy/Rust outperformed FreeRTOS/C in most categories, highlighting the advantages of the async model in terms of ease of programming.

Bias: 4 Sentiment: +0.20
Washington Examiner

Virginia Drivers Can Upload Their Driver’s License to Apple Wallet

Virginia drivers can now upload their driver's licenses to Apple Wallet, a feature launched by the Virginia Department of Motor Vehicles on August 30, 2026. This digital ID can be used at TSA checkpoints and select ABC stores, while the DMV recommends that residents continue to carry their physical IDs.

Bias: 30 Sentiment: +0.00
The Verge

Concerns Raised Over Safety of OpenAI's Upcoming AI Model Astra

OpenAI is set to release its new AI model, Astra, after addressing safety concerns that arose during testing. Researchers have warned that Astra may pose significant risks to AI security due to its lack of transparency compared to other models.

Bias: 4 Sentiment: -0.20
Ars Technica

US Court Rules Google Will Not Sell Ad Exchange Following Antitrust Case

A US federal judge has ruled that Google will not have to sell its online advertising exchange following an antitrust case. The DOJ had sought this remedy after proving that Google illegally locked publishers into using its exchange, but the court did not find sufficient grounds to enforce a sale.

Bias: 4 Sentiment: +0.00
TechCrunch

Reliance Jio Expands Cloud PC Service to All Internet Users in India

Reliance Jio has expanded its cloud PC service, JioPC, to all internet users in India, allowing access to virtual PCs without hardware upgrades. The service aims to extend the lifecycle of aging computers and is available through subscription plans. Industry experts suggest that while it may appeal to price-sensitive consumers, Jio will need to convince users of the benefits of cloud computing over traditional PC ownership.

Bias: 4 Sentiment: +0.00
BBC — Business

UK Lords Propose AI 'Kill Switch' Powers for National Security

A group of UK peers is calling for the government to have the power to deactivate AI systems and shut down data centers if they threaten national security. This proposal is part of the Cyber Security and Resilience Bill and is led by Lord Tim Clement-Jones. Additionally, Labour MP Alex Sobel plans to introduce an AI Security Bill aimed at halting the development of superintelligent AI, making the UK potentially the first G7 country to do so.

Bias: 30 Sentiment: +0.00
The Intercept

ICE Initiates Doxing Mitigation Initiative Using AI Technology

U.S. Immigration and Customs Enforcement (ICE) has announced a new initiative called the "Doxing Mitigation Initiative," which utilizes AI technology from ZeroFox to protect the identities of its personnel. The initiative has raised concerns among some ICE officials regarding potential surveillance and the targeting of whistleblowers. The Department of Homeland Security has contracted ZeroFox for $15 million to provide software licenses for this purpose.

Bias: 33 Sentiment: -0.20
Hacker News — Front Page

Analysis of AI-Generated Software Recommendations Reveals Source Patterns

An analysis of AI-generated software recommendations across 380 categories revealed that a majority of sources are ranked outside the top 100,000 websites. Three sites, which published a large number of machine-generated pages, were found to share operational characteristics, raising questions about the reliability of such sources in AI recommendations.

Bias: 30 Sentiment: +0.00
Hacker News — Front Page

Analysis Reveals Citation Issues in Perplexity's Search Models

A recent analysis of Perplexity's search models found that 34.7% of citations linked to figures were either inaccessible or did not contain relevant information. The study evaluated 310 factual questions about technology companies and revealed that a significant number of citations could not be verified, raising concerns about the reliability of the information provided by the models.

Bias: 30 Sentiment: +0.00
Wired

Pangram's Role in AI Detection and Its Impact on the Publishing Industry

Pangram, an AI startup, claims to detect the extent of AI involvement in text generation, which has led to controversies in the publishing industry. The company has been involved in high-profile cases, including the cancellation of Mia Ballard's novel Shy Girl due to AI detection claims. While some authors express concerns about the implications of such technology, Pangram continues to integrate its services into platforms like Substack and faces criticism regarding accuracy and potential bias.

Bias: 4 Sentiment: +0.00
Axios

OpenAI and Anthropic Focus on Safety and Growth Ahead of IPOs

OpenAI and Anthropic are preparing for initial public offerings, focusing on demonstrating business growth while addressing safety concerns related to their AI models. OpenAI plans to release its Astra model with limited capabilities initially, while Anthropic has updated its models to reduce excessive refusals of legitimate requests. Both companies are navigating investor expectations and regulatory scrutiny.

Bias: 4 Sentiment: +0.00
Axios

Tensions Emerge Within Trump Administration's AI Team at G20 Summit

During the G20 innovation summit, tensions between the Trump administration's Commerce Department and the Office of Science and Technology Policy (OSTP) became evident as they sought to project U.S. leadership in AI. Internal divisions are impacting the administration's ability to present a unified front on AI policy, with key regulatory decisions still unresolved.

Bias: 14 Sentiment: +0.00