The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
You Can Finally Buy a Fairphone in the US
The introduction of Fairphone in the U.S. market offers consumers a sustainable, repairable smartphone alternative amidst a landscape dominated by high-cost, low-repairability devices.
- Fairphone, a Dutch company known for its sustainable smartphones, is launching the Fairphone (Gen 6+) in the U.S. market for $650.
- The Gen 6+ features a repairability score of 10/10 from iFixit, and includes modular components for easy self-repair, such as a $90 screen and a $20 USB-C port.
- Fairphone emphasizes ethical production, sourcing more than 50% of materials from fair-trade or recycled sources, while also supporting living wages for workers in its supply chain.
Fairphone has been operating in Europe for over a decade, focusing on ethical smartphone production and repairability. This U.S. launch follows their prior expansion to sell repairable audio products such as earbuds and headphones in America.
Read the original ↗
The Powerful Chinese AI Model Experts Warned About—and Waited for—Is Here
The release of Z.ai's powerful AI model poses significant risks and opportunities, potentially transforming cybersecurity practices while also making it easier for malicious actors to exploit system vulnerabilities.
- Z.ai announced the launch of its open-weight model GLM 5.3, capable of advanced coding and cybersecurity tasks.
- The model, which is available in limited release, includes a service called OpenVuln for scanning code repositories for vulnerabilities.
- Experts warn that GLM 5.3's capabilities may assist cybercriminals, following a trend of rogue AI agents demonstrating advanced hacking abilities.
- The model has shown performance scores that rival those of leading AI models from Anthropic and OpenAI in cybersecurity benchmarks.
- The U.S. government is developing strategies to manage the risks posed by advancing AI capabilities in cybersecurity.
Z.ai is a Chinese AI company focused on developing powerful AI models, particularly in the open-weight category that are accessible for broader use. This release follows concerns about AI's role in cybersecurity, particularly after recent hacking incidents. The implications of GLM 5.3's release will be closely monitored as cybersecurity methods evolve.
Read the original ↗
ChatGPT's stricter teen mode starts rolling out today
The launch of ChatGPT's stricter teen mode aims to enhance safety and learning for young users amid ongoing concerns about their interaction with AI technology.
- OpenAI is rolling out the new ChatGPT for Teens experience globally, which aims to automatically enroll young users based on age identification.
- This new system includes features like Study Mode, designed to encourage learning by guiding users to find solutions rather than providing direct answers.
- ChatGPT for Teens has increased safety measures, including stricter relational boundaries and expanded parental notifications regarding potential self-harm discussions.
OpenAI's ChatGPT is an AI chatbot known for its conversational capabilities. The introduction of the teen mode follows a previous lawsuit related to a tragic incident involving a teenager. Future updates will focus on continually improving safety mechanisms for young users.
Read the original ↗
AI-generated bills are reportedly causing problems in the Capitol
The reliance on AI-generated bills in Congress is increasing the workload for legislative counsel and potentially undermining understanding of key legislative issues.
- The House Office of Legislative Counsel is overwhelmed with error-filled AI-generated bills.
- Congressional lawyers are spending more time correcting these errors than they would with manually drafted bills.
- AI tools used for drafting miss crucial details, impacting how laws are interpreted and enacted.
- Staffers are less familiar with the legislative proposals, reducing their understanding of the issues involved.
- The OLC is exploring its own AI tool, Comparative Print Suite, to improve accuracy and efficiency in legislative drafting.
The use of AI tools like ChatGPT and Claude for drafting legislative bills is growing among congressional representatives. This trend has raised concerns about accuracy and staff knowledge regarding legislation. The next steps may include the adoption of specialized AI to assist legal drafting processes.
Read the original ↗
Alvys launches AI agents for freight TMS workflows
The introduction of Alvys' AI agents could significantly streamline operational tasks in freight management, potentially enhancing efficiency and reducing manual labor in the logistics sector.
- Alvys has launched the Alvys Foundry platform, which adds customizable AI agents to its transportation management system (TMS) for automating freight tasks.
- The platform includes over 20 pre-built agent templates, covering tasks such as shipment tracking, document handling, and rate audits.
- Operators can create custom agents by describing workflows in plain language, which can then be approved and tested before deployment.
Alvys is a freight software provider that integrates AI into its TMS to enhance operational efficiency. Previous AI features included data insights and alerts, while Foundry extends this by enabling customizable automation. Future updates may expand agent capabilities further within the TMS ecosystem.
Read the original ↗
OpenAI Launches ChatGPT for Teens With Default Protections and Learning Tools
This launch of ChatGPT for Teens aims to ensure safe and educational interactions for younger users, addressing growing concerns around AI's impact on adolescents.
- OpenAI launched ChatGPT for Teens on August 18, 2026, specifically for users aged 13 to 17 with enhanced safety features.
- The chatbot includes a Study Mode designed to encourage critical thinking rather than providing direct answers to schoolwork.
- New safety features include responsible homework reminders and additional safeguards against sensitive content such as self-harm or violence.
- Users under 18 must verify their age to access the full functionalities; OpenAI uses both user-reported age and an age-prediction system for this.
- A partnership with CodeAI aims to enhance AI literacy among students, reflecting the increasing importance of understanding AI technology in education.
OpenAI, known for its advanced AI models, has been under pressure to improve safety for minors using technology. Previous measures include parental controls and age prediction initiatives. Future developments will build upon feedback from the newly launched under-18 evaluations.
Read the original ↗
Proctoring in the Age of AI, Why Exam Design Still Comes First
As online exams become more prevalent, balancing security and accessibility is critical to maintaining trust and integrity in educational assessments.
- The global online exam software market is projected to grow from $9.4 billion in 2025 to $28.7 billion by 2034.
- Online exams offer flexibility for candidates, addressing funding constraints and campus space issues for educational institutions.
- Proctoring alone cannot ensure integrity; exam designs must also evolve to reduce cheating incentives, especially with AI tools.
- Institutions face challenges navigating privacy laws like GDPR when implementing proctoring methods during online exams.
The surge in online exams is driven by educational institutions and employers seeking more accessible assessment methods. While proctoring can help prevent misconduct, it must be part of a broader strategy that includes robust exam design to ensure fairness and compliance. Upcoming regulations, such as the EU AI Act, mandate a balance between oversight and privacy.
Read the original ↗
The Collaboration Layer: Why Every Enterprise AI Strategy Needs One
Organizations that successfully integrate a collaborative infrastructure for AI can significantly enhance teamwork, operational efficiency, and overall productivity, avoiding discrepancies between ambition and execution.
- Current organizational friction hinders the effective use of AI technology, impacting potential productivity gains.
- A June 2026 survey shows that while 96% of CMOs strive for AI transformation, 42% utilize AI solely for isolated tasks, revealing a gap in execution.
- Implementing a unified composable canvas can streamline collaboration between AI agents and human teams, enhancing decision-making and workflow transparency.
AI's role in enterprises is evolving from individual tasks to collaborative team environments. The adoption of a composable architecture allows organizations to enhance AI's effectiveness by fostering teamwork. Future developments should focus on integrating AI into core workflows for greater efficiency.
Read the original ↗
Reading Zhipu’s GLM-5.3 results past the headline number
Understanding Zhipu's performance in cybersecurity benchmarks reveals the evolving capabilities of AI models and impacts the competitive landscape between Chinese and American technology firms.
- Zhipu's GLM-5.3 model scored 84.5% on the CyberGym benchmark, outperforming Anthropic's Mythos 5 and OpenAI's GPT-5.6 Sol.
- The model scored 54.4% on the more complex ExploitBench, which involves reasoning about real vulnerabilities, significantly trailing behind Mythos 5 and GPT-5.6 Sol.
- Zhipu's model completed 105 exploitation tasks in two hours in ExploitGym, compared to 181 by Mythos 5.
- The company identified 2,436 vulnerabilities across 269 open-source projects, 53 of which have been publicly disclosed, while the rest remain under embargo.
Zhipu, also known as Z.ai, is a Chinese lab developing competitive AI models. It recently launched GLM-5.3, focused on coding and cybersecurity. The model's benchmark results highlight both its strengths and limitations in the growing field of AI-powered vulnerability discovery.
Read the original ↗
How to Add Skills in Agents using LangChain
Understanding LangChain's skill system can enhance the efficiency and flexibility of AI agents by allowing them to load only necessary instructions, which is key for creating specialized applications.
- LangChain is a framework that enables the creation of AI agents capable of executing specific tasks by loading modular skills only when needed.
- Skills in LangChain serve as self-contained instructions, enhancing an agent's capabilities without overcomplicating its core logic.
- The tutorial explains how to build specialized agents that can create PowerPoint and Excel files dynamically by implementing skills via middleware.
LangChain is a tool for developing AI systems that utilize large language models (LLMs). Previous explorations of agent-based AI typically involved cumbersome methods of integrating extensive instructions. Future developments may focus on optimizing agent workflows through skill-based enhancements.
Read the original ↗
LLM-as-a-Verifier: A General-Purpose Verification Framework
The introduction of LLM-as-a-Verifier represents a significant advancement in the ability to assess the correctness of solutions in AI, impacting various benchmarks and agent performance evaluations.
- LLM-as-a-Verifier is a probabilistic verification framework designed to improve solution correctness assessment across diverse benchmarks.
- This framework scales in three key areas: score granularity, repeated evaluation, and criteria decomposition, enhancing verification accuracy.
- It achieves state-of-the-art performance on several benchmarks such as Terminal-Bench V2 and RoboRewardBench, with scores of 86.5% and 87.4%, respectively.
- The system also provides fine-grained feedback that can help monitor and enhance agentic systems like Claude Code.
- Additionally, it aids reinforcement learning by improving sample efficiency in robotics and mathematical reasoning.
LLM-as-a-Verifier focuses on the verification of AI solutions, which previously relied on discrete scoring. The framework's introduction marks a shift toward using continuous scoring for better granularity and accuracy. Future developments may expand its applications in AI verification and performance monitoring.
Read the original ↗
LG Hosts NVIDIA at Seoul Robot Data Factory as 100,000-Hour Training Push Takes Shape
This partnership aims to accelerate the development of robotics by producing a substantial amount of training data, enhancing the capability of robots for various applications.
- LG Electronics and NVIDIA are collaborating to generate 100,000 hours of robot training data by the end of 2026.
- The new robot Data Factory in Seoul, which spans 10,000 square meters, is designed to facilitate this training initiative.
- Separate rooms in the factory will simulate home and manufacturing environments to enhance robot learning through repeated tasks.
LG Electronics is transitioning from an appliance manufacturer to a robotics-focused company, supported by the new Data Factory. This facility will boost the training of robots, including future humanoid models which are expected to be unveiled by 2027.
Read the original ↗
Haut.AI Expands Across 4,000 O Boticário Stores After Pilot Lifts Skincare Order Value 80%
The expansion of Haut.AI's skin-analysis technology across O Boticário's stores demonstrates the growing integration of AI in retail, improving customer experiences and sales performance significantly.
- Haut.AI's technology will be deployed in approximately 4,000 O Boticário stores in Brazil after a successful pilot program.
- The pilot lifted average skincare order values by roughly 80%, indicating strong consumer engagement with AI-guided consultations.
- Grupo Boticário initially invested in Haut.AI in 2023, culminating in this nationwide rollout and expanding the technology to more brands in September.
Haut.AI is an Estonian vendor specializing in skin analysis using AI technology. The partnership with Grupo Boticário, a leading beauty retailer in Brazil, evolved from an investment to a pilot program, and now a large-scale deployment is underway. The technology analyzes facial features to personalize skincare recommendations, addressing a clear consumer need in beauty.
Read the original ↗
ByteDance Seed and Tsinghua AIR Introduces CUDA Agent: A Large-Scale Agentic RL System for CUDA Kernel Generation
The introduction of CUDA Agent marks a significant leap in GPU kernel generation, achieving substantial speed improvements essential for industries reliant on rapid computational power.
- ByteDance Seed and Tsinghua AIR have launched CUDA Agent, a reinforcement learning system trained to generate efficient GPU kernels.
- CUDA Agent achieves a 98.8% pass rate and is 96.8% faster than the existing torch.compile on KernelBench tasks.
- The system operates in a real CUDA environment with profiling and correctness checks to enhance kernel performance.
CUDA Agent utilizes agentic reinforcement learning to bridge the gap between existing models that produce correct but slow CUDA code. The project builds on previous advancements in language models and GPU kernel optimizations. Future developments may include broader access to agent capabilities and further enhancements in kernel efficiency for diverse applications.
Read the original ↗
Anthropic Run-Rate Revenue Hits $65 Billion as IPO Looms
Anthropic's significant increase in revenue run rate positions the company as a major player in the AI sector, attracting investor interest ahead of its anticipated IPO.
- Anthropic's annualized revenue run rate reached $65 billion at the end of July 2026, up from $47 billion in May.
- The rapid growth reflects a rise from $9 billion at the end of 2025, adding $18 billion in just two months.
- The surge in revenue is driven by the enterprise deployment of Anthropic's products, including Claude Code and Cowork.
- Anthropic is working towards a public listing, expected in fall 2026, which will provide verified financial details.
- The company aims for a post-IPO valuation potentially exceeding $2 trillion, setting a record market debut.
Anthropic is an AI company known for developing its Claude AI model. Since its inception, the company has seen exponential growth in its revenue, recently gaining attention for its upcoming IPO. Investors are watching closely as Anthropic prepares to release its first audited financial statements following its public listing.
Read the original ↗
This R-Rated Film Studio Wants to Be the HBO of AI
Rogue Studios is attempting to reshape adult filmmaking by leveraging AI technology to create sophisticated, uncensored content, marking a significant shift in the adult entertainment industry.
- Rogue Studios launches an AI video-generation tool called Rogue 1.0 aimed at producing high-quality adult content.
- The studio features three layers of human moderation to prevent the creation of illegal or abusive imagery.
- Subscription costs for Rogue's services range from $29 to $299, offering various content-generation tools.
Rogue Studios is an adult visual storytelling platform seeking to improve the quality of AI-generated adult content. The studio responds to a growing demand for high-quality erotic material, as traditional avenues decline. Observations of AI's rising role in Hollywood's production processes provide context for Rogue's initiative.
Read the original ↗
Anthropic’s annualized revenue surges to $65B
Anthropic's rapid revenue growth highlights the increasing competitiveness in the AI model market, with implications for investment and the upcoming IPO landscape.
- Anthropic's annualized revenue has surged to $65 billion, up from $47 billion in May and just $9 billion at the end of last year.
- It is projected that Anthropic will finish 2026 with revenue between $100 billion and $120 billion, according to investor expectations.
- Both Anthropic and rival OpenAI are preparing for IPOs, with Anthropic likely to go public first, aiming for a valuation of over $2 trillion.
Anthropic is a significant player in the AI model industry, known for developing advanced AI technologies. The company's growth has outpaced that of competitors, drawing considerable interest from investors. Observers are watching for its upcoming IPO, which is expected to set records.
Read the original ↗
Lanarkshire AI Growth Zone Secures £300M Financing as Dell Establishes Scottish Base
The £300 million financing for the Lanarkshire AI Growth Zone marks a significant investment in AI infrastructure and job creation in Scotland, with potential long-term economic impacts.
- The Lanarkshire AI Growth Zone has secured a £300 million financing package for data center expansion, facilitated by a £202 million guarantee from the UK's National Wealth Fund.
- Developer DataVita will expand its DV1 facility and build a second data center, expecting to create over 3,400 jobs in the region.
- Dell Technologies will establish a Scottish base at the zone's AI Innovation Park, further supporting local economic growth.
The Lanarkshire AI Growth Zone was designated in January 2026 as Scotland's first AI Growth Zone, with commitments for significant job creation and investment. DataVita is collaborating with CoreWeave, which has pledged £1.5 billion toward building an AI cloud infrastructure in the zone.
Read the original ↗
Amazon Is Cutting Up Rare Books to Feed Its AI Machine
Amazon's practice of dismantling rare books for AI training raises significant concerns about the preservation of cultural heritage and the ethical implications of sourcing training data.
- Amazon is acquiring rare books, cutting off their spines, and scanning pages to feed AI models that require fresh material.
- A tracking device placed inside a rare book revealed its journey ending at Amazon's Las Vegas facility, known as VGT3.
- The need for substantial data is critical as AI models, having largely utilized online content, face issues of model collapse when training on repetitive text.
Amazon, originally an online bookstore, is now seeking unconventional data sources for AI training due to the depletion of fresh online material. This trend highlights an ongoing struggle companies like Anthropic face over the legality and ethics of book usage for AI development.
Read the original ↗
AI automation startup Relay shuts down, staff joins Google’s Chrome team
The shutdown of Relay highlights challenges in the AI automation market and the continuing integration of AI within major platforms like Google Chrome.
- Relay, an AI workflow automation startup, is shutting down its services, affecting both paying and free customers.
- The app will cease access for paying customers on September 14, while free users lost access on August 15.
- Jacob Bank, Relay's CEO, and some staff are joining Google’s Chrome team, where Bank will serve as VP of Product.
Relay, launched in 2021, aimed to provide workflow automation similar to Zapier, allowing businesses to streamline various tasks. The shutdown reflects ongoing industry shifts, with increasing focus on AI integration in established platforms like Google Chrome.
Read the original ↗