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
#507 Anthropic Says Alibaba Stole 29 Million Conversations With Claude
This case highlights the ongoing issues of intellectual property theft in AI, particularly as major tech firms compete for advanced capabilities, which could impact future innovations and regulations.
- Anthropic has accused Alibaba of employing 25,000 fake accounts to extract nearly 29 million conversations from its AI, Claude.
- The allegations describe the use of nefarious methods termed 'adversarial distillation' to harvest Claude's key functionalities.
- This incident marks the first time Anthropic has publicly identified a major Chinese tech firm as the source of a model-theft campaign.
Anthropic is an AI research laboratory focused on developing advanced AI systems. The incident underscores the competitive and aggressive landscape of AI development, where companies are vying for top talent and intellectual property. Future developments should be monitored as industry regulations and standards emerge in response to these practices.
Read the original ↗
IBM claims world’s first sub-1 nanometer chip technology
IBM's new chip technology could significantly improve the performance and energy efficiency of AI data centers, influencing future computing capabilities and design standards in the semiconductor industry.
- IBM has developed what it claims to be the world's first sub-1 nanometer chip technology, enabling nearly 100 billion transistors on a chip the size of a human fingernail.
- The new nanostack architecture can potentially provide 50% higher computing performance and 70% greater energy efficiency compared to IBM's previous 2-nanometer node chips.
- This chip architecture incorporates a staggered layout of vertically stacked transistors, addressing physical scaling limits in modern chip design.
- IBM's advancements could lead to substantial improvements in static random-access memory (SRAM), which is critical for AI applications, with projections of a 40% scaling improvement.
- Commercial production of chips using this new technology may begin within five to ten years, as IBM seeks partnerships for manufacturing.
IBM's new chip technology is part of ongoing efforts in semiconductor research aiming to increase transistor density and performance. Previous advancements include the 2-nanometer node introduced in 2021, and the new architecture builds on earlier concepts like nanosheet transistors. The semiconductor landscape is marked by rapid advancements, with companies exploring various strategies for improved chip efficiency and capability.
Read the original ↗
AI is Reshaping Trade Compliance, and Most Companies Aren’t Ready
As trade compliance becomes increasingly complex due to AI integration, organizations risk falling behind if they do not adapt their strategies and skills in this evolving landscape.
- Trade compliance is now central to operations, influenced by tariff volatility and enforcement of forced labor laws.
- AI is transforming product classification and risk monitoring in trade compliance, necessitating new skill sets for professionals.
- Companies face challenges in maintaining consistent product data across various disconnected systems, which increases compliance risks.
Trade compliance ensures businesses adhere to laws when trading internationally. Historically overlooked, recent geopolitical events and AI advancements are pushing companies to modernize compliance practices. Moving forward, organizations need to enhance collaboration and data literacy to manage compliance effectively under evolving regulations.
Read the original ↗
Why AI Isn’t Paying Off for the Middle Market
The low AI Adoption Index in the middle market highlights a significant operational challenge, indicating that many companies struggle to derive meaningful value from their AI initiatives despite making investments.
- CBIZ's AI Adoption Index for the middle market is only 35 out of 100, indicating 'fragmented adoption' without clear strategy.
- Nearly half of surveyed organizations cite a lack of internal AI expertise as their primary barrier to effective AI implementation.
- Success requires business leaders' active involvement in AI projects rather than treating them as merely IT functions.
The middle market refers to companies with annual revenues typically between $10 million and $1 billion, which are increasingly pursuing AI for operational improvements. Despite recent investments, many firms are facing challenges in translating AI initiatives into measurable business outcomes.
Read the original ↗
The Self-Improving Loop in AI Agents: Architecture, Benefits, and H...
The self-improving loop in AI agents enhances performance by enabling continuous learning from past tasks, reducing repetitive errors and improving overall task efficiency.
- Self-improving agents can learn from their results, applying lessons to improve future performance.
- Traditional AI agents often forget past outcomes, leading to repeated mistakes without learning from errors.
- The self-improving loop incorporates feedback mechanisms to foster ongoing improvement in agent capabilities.
AI agents, traditionally used in various applications, often lack long-term learning capabilities, resulting in repetitive errors. The self-improving loop introduces a feedback mechanism that allows agents to learn from past experiences, significantly enhancing their performance. Understanding this architecture can help improve future AI development processes.
Read the original ↗
AI Will Not Take Engineering Jobs
The increasing demand for engineers despite advancements in AI suggests a significant shift in workforce dynamics and the potential for job growth in the tech sector.
- Recent data from SignalFire shows that while overall hiring at large tech companies fell by 25%, engineering roles only dropped by 11%.
- Engineers comprised 55% of all new hires at major tech firms in 2025, up from 46% in 2019.
- Early-stage startups hired 7% more engineers in 2025 compared to 2019, indicating robust demand in the startup sector.
SignalFire is a venture firm analyzing job trends across millions of workers and companies. Despite predictions of AI replacing jobs, evidence shows engineers are increasingly in demand as AI tools enhance productivity and project possibilities.
Read the original ↗
OpenAI's free GPT-5.5 model makes ChatGPT better at understanding context
OpenAI's upgrade to GPT-5.5 Instant enhances ChatGPT's ability to handle complex questions and provide relevant responses, improving user experience and interaction quality.
- OpenAI upgraded its GPT-5.5 Instant model to improve context understanding and adaptability in ChatGPT.
- The model now produces 52.5% fewer hallucinations and 37.3% fewer factual errors compared to previous versions.
- GPT-5.5 Instant can now better process complex inquiries and carry context across multiple exchanges, improving user relevance in responses.
- The model also offers improved location-based recommendations and can supply relevant images, enhancing overall interaction.
- These updates aim to make responses feel more natural and cohesive, reducing the templated look of previous replies.
OpenAI's GPT-5.5 Instant is the default model for its ChatGPT chatbot, designed to simulate human-like conversations. The previous version faced challenges with factual accuracy and contextual understanding, prompting ongoing upgrades to enhance user experience. Users can expect a more nuanced interaction as the technology continues to evolve.
Read the original ↗
The math behind the OpenAI Jalapeño chip
Understanding the economic implications of the OpenAI Jalapeño chip could influence future AI development and deployment strategies across the industry.
- The article discusses the mathematics underlying the OpenAI Jalapeño chip.
- It explores inference economics related to this new chip.
The OpenAI Jalapeño chip is designed to enhance AI processing capabilities. This chip represents a significant advancement in AI hardware aimed at optimizing inference, which is the application of a model to new data. Observers should monitor how this chip influences AI performance and cost-effectiveness moving forward.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Baidu Releases Unlimited OCR, a 3B Model That Keeps the KV Cache Flat for Long-Document Parsing
The advancement of Baidu's Unlimited OCR may enhance the efficiency of processing long documents, which is crucial for various applications involving extensive text analysis.
- Baidu has released Unlimited OCR, a model designed for optical character recognition (OCR).
- The new model features 3 billion parameters and maintains a flat key-value (KV) cache during long-document parsing.
Baidu is a Chinese technology company specializing in AI and internet services. OCR technology is significant for converting different types of documents into editable and searchable data. The release of Unlimited OCR reflects ongoing advancements in AI models for text processing.
Based on the headline only — the full article could not be fetched.
Read the original ↗
EXL to Acquire iMerit in Up to $310 Million Deal to Expand Enterprise AI and Foundation Model Capabilities
This acquisition enhances EXL's capabilities in AI infrastructure, helping organizations integrate advanced AI models and training data crucial for industries like healthcare and autonomous systems.
- EXL has signed a definitive agreement to acquire iMerit for a transaction valued at up to $310 million.
- This acquisition aims to strengthen EXL’s position in enterprise AI by improving foundation model development and multimodal data operations.
- The deal is expected to close during the third quarter of 2026, pending regulatory approvals.
- iMerit specializes in high-quality training data and human evaluation for AI models used in various sectors, including healthcare and robotics.
- EXL plans to integrate iMerit’s Ango platform into its AI offerings, enhancing its technology stack for continuous model improvement.
EXL, founded in 1999 and originally a business process outsourcing company, has transitioned into an AI-focused enterprise technology firm. The company aims to help organizations apply AI effectively across their operations, moving from isolated projects to integrated solutions. The acquisition of iMerit is part of EXL's strategy to enhance its AI capabilities amid the growing need for robust AI training and evaluation infrastructure.
Read the original ↗
The Best Prime Day Deals on LED Masks and Hair Growth Devices We've Tested
These discounts on LED masks and hair growth devices during Prime Day present consumers with affordable options to invest in health and beauty technologies that can support skin and hair wellness.
- Amazon Prime Day offers significant discounts on red light therapy devices, benefitting consumers looking to enhance skincare and hair restoration routines.
- The Dr. Dennis Gross DRx SpectraLite Faceware Pro is available for $364, with a $91 discount, making it a compelling option for skincare enthusiasts.
- CurrentBody LED Hair Growth Helmet, popular for hair restoration, is discounted to $770, saving users $90 on a hands-free treatment option.
- The Laduora Lumeo SkinLift 4D Current Red Light Therapy Kit is priced at $161, offering a 4-in-1 solution for skincare with a $54 discount.
Red light therapy devices utilize specific light wavelengths to promote skin health and hair restoration. Prices for these gadgets can be high, making Prime Day an ideal time for consumers to consider purchases. Notable products and discounts can help buyers navigate their choices in this growing market.
Read the original ↗
A24 Knows You’re Mad About the Google AI Collab
The backlash against A24's investment in AI raises questions about the impact of artificial intelligence on creativity and the film industry.
- A24 has announced a $75 million research partnership with Google DeepMind to create new filmmaking tools through A24 Labs.
- The announcement follows A24's recent success with the horror film 'Backrooms,' which has grossed over $300 million globally.
- Fans have reacted negatively to the partnership, expressing concerns on social media about its implications for A24's artistic integrity.
A24 is an influential indie film studio known for critically acclaimed films. The partnership with DeepMind exemplifies the growing intersection between technology companies and Hollywood, stirring debate over AI's role in creative industries.
Read the original ↗
Here’s How to Opt Out of Google Search’s New AI Data Training Feature
Google's new data training feature affects user privacy, requiring proactive steps to opt out, which could influence public perception and usage patterns of its services.
- Google has introduced a new setting called Search Services History that retains user-uploaded media for AI model training by default.
- Users can opt out of this feature by visiting Google's My Activity page and unchecking the box next to Save media.
- Media saved includes images, audio, and video from interactions like Google Lens and Google Translate, and may be stored for up to four years.
- Experts express concern that this 'opt-out' system places an undue burden on users to manage their privacy settings effectively.
- Critics argue that data collection practices may lead to feelings of powerlessness among users regarding their personal information.
Google is a major technology company that offers a range of online services, including search engines and translation tools. Previous discussions have raised concerns about how user data is used for improving AI models. Future scrutiny will likely focus on user engagement with these privacy settings and related features in Google's ecosystem.
Read the original ↗
OpenAI and Broadcom announce chip designed for LLM inference at scale
The announcement of the Jalapeño chip signifies a strategic move by OpenAI and Broadcom to enhance performance and reduce reliance on external chip suppliers in the AI sector.
- OpenAI and Broadcom have collaborated to develop the Jalapeño chip, specifically designed for large language model (LLM) inference in data centers.
- The chip's design was informed by OpenAI's roadmap and took nine months to develop from scratch as an Application-Specific Integrated Circuit (ASIC).
- The Jalapeño chip is expected to outperform current LLM inference systems in energy efficiency, although detailed performance metrics will be released later.
- Both companies aim to have Jalapeño chips operational in data centers by the end of this year.
OpenAI, known for its AI models like ChatGPT, partnered with Broadcom, a major semiconductor supplier, to create specialized chips for advanced AI workloads. This development comes amidst a surge in demand for AI computing resources, with companies seeking custom silicon solutions to enhance performance.
Read the original ↗
Spacelift’s The AI Readiness Gap Report Reveals a Growing Divide Between AI Adoption and Infrastructure Governance
As AI adoption outpaces infrastructure governance, the increasing number and severity of AI-related incidents underscores an urgent need for organizations to address their governance frameworks.
- The Spacelift report reveals a significant gap between AI adoption and the governance of infrastructure within organizations.
- While 89% of organizations report increased developer velocity due to AI, only 25% prioritize AI governance strategically.
- 93% of organizations experienced at least one AI-related infrastructure incident in the past year, highlighting governance challenges.
The report examines the impact of AI on infrastructure management as many organizations fail to align their governance practices with accelerating AI adoption. It introduces the AI Maturity Index, categorizing organizations based on their governance and integration of AI. These findings indicate a pressing need for stronger governance frameworks to manage emerging AI risks effectively.
Read the original ↗
This Is the Only Roborock Robot Vacuum Sale Worth Shopping
Consumer interest in robotic vacuum technology continues to grow, making insights about effective models particularly relevant for budgeting and home cleaning efficiency.
- Roborock offers a variety of robot vacuum models, including the Saros series and Qrevo Curv series, available for sale during Prime Day.
- The recommended model is the Roborock Qrevo Curv 2 Flow, noted for its advantageous price and effective functionality.
- The Qrevo Curv 2 Flow has 20,000 Pa of suction and features a self-cleaning spinning roller mop, making it suitable for homes with hard floors.
Roborock is a brand renowned for its advanced robotic vacuum cleaners. The current sale coincides with Amazon's Prime Day promotion, offering discounts on selected models, which highlights consumer interest in such technologies.
Read the original ↗
Jalapeño is the first AI chip from OpenAI and Broadcom
The introduction of the Jalapeño chip signifies a pivotal step in enhancing AI efficiency, potentially improving the performance of large language models and their applications.
- OpenAI and Broadcom have announced the Jalapeño chip, their first joint AI accelerator, with a focus on large language model (LLM) inference.
- This collaboration was initiated in October 2025, aiming for custom chip production to optimize performance in AI tasks.
- Jalapeño is reported to offer better performance per watt compared to existing chip technologies, although final performance tests are ongoing.
- The initial deployment of Jalapeño in data centers is scheduled to begin in late 2026.
- The chip was designed and developed in just nine months, reflecting a close collaboration between OpenAI and Broadcom.
OpenAI is a leader in artificial intelligence, known for its large language models like ChatGPT. Previously, it relied on external chips for AI processing, but now, it aims to produce its own to enhance efficiency. Future developments will likely focus on further improving AI capabilities and integration with existing systems.
Read the original ↗
Companies are scrambling to stop employees from maxing out AI budgets with small tasks
Companies are reevaluating their AI spending priorities in light of rising costs, which could reshape their operational strategies and employee expectations.
- Many companies are restricting their employees' use of AI for minor tasks to manage escalating costs.
- Accenture is scrutinizing token expenditures after observing significant financial losses from excessive use on basic functions.
- There's an ongoing discourse about the effectiveness of AI spending at the executive level, particularly among CFOs and COOs.
The AI industry has shifted from aggressive budget utilization, termed 'tokenmaxxing', to a more cautious approach due to rising costs and limited returns. Companies are reconsidering how they deploy AI resources amidst concerns about financial effectiveness and value.
Read the original ↗
AI researchers continue to leave Google for its rivals
The ongoing exodus of AI researchers from Google to rivals like Anthropic and OpenAI threatens Google's competitive position in artificial intelligence innovation.
- Top AI researchers Jonas Adler and Alexander Pritzel are leaving Google for Anthropic, contributing to a pattern of departures.
- Noam Shazeer, a prominent AI researcher, also recently left Google for OpenAI after a long tenure.
- John Jumper, director at Google DeepMind, announced his transition to Anthropic, highlighting a significant loss for the company.
Google has faced increasing talent drain as prominent AI researchers seek opportunities at competing companies like OpenAI and Anthropic, which are preparing for potential public offerings. The trend raises concerns about Google's ability to maintain its leadership in AI development.
Read the original ↗
AI was supposed to kill engineering jobs, but new data suggests they’re the most resilient
The resilience of engineering jobs amid AI advancements suggests that technological progress may not eliminate these positions but instead increase their demand and productivity.
- Research from SignalFire indicates that engineering roles experienced only an 11% decline in hiring compared to a 25% decline overall in large tech companies since 2019.
- Engineers accounted for 55% of new hires among 12 major tech firms in 2025, up from 46% in 2019, highlighting growing demand for engineering talent.
- Early-stage startups hired 7% more engineers in 2025 than in 2019, reflecting continued investment in engineering despite tech industry layoffs.
SignalFire is a venture firm that analyzes workforce trends across millions of employees and companies. Previous concerns suggested AI might endanger software engineering roles, but recent data shows engineering demand is increasing. Stakeholders will monitor the ongoing role of AI in workforce dynamics and its impact on hiring trends.
Read the original ↗