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
A Mysterious Roman Artifact Covered in Odd Markings Puzzled Researchers for More Than a Century—Now AI Has Cracked Its Ancient Code
This story highlights the intersection of artificial intelligence and archaeology, revealing how advanced technology can reinterpret ancient artifacts and reshape our understanding of history.
- Researchers have identified a Roman artifact known as Object 04433, featuring unusual markings, as potentially linked to ancient blockade games.
- Utilizing AI simulations, researchers tested 130 game configurations to match the wear patterns on the artifact's surface.
- The findings suggest that blockade games may have existed in ancient Rome, pushing back their historical timeline by several centuries.
Object 04433 is a Roman artifact housed in the Roman Museum in the Netherlands, uncovered over a century ago and previously regarded as a possible gaming board. The recent application of artificial intelligence in archaeological research demonstrates its potential to unlock new insights into historical artifacts and practices.
Read the original ↗
Uber’s new autonomous vehicle division is about survival and opportunity
Uber Autonomous Solutions aims to position Uber as a key player in the commercialization of autonomous vehicles, potentially reshaping transportation and logistics industries.
- Uber has launched a new division called Uber Autonomous Solutions to support robotaxi, self-driving truck, and delivery robot operations.
- The company has formed partnerships with nearly two dozen autonomous vehicle technology companies, including Nuro and WeRide.
- Uber plans to assist partners in scaling robotaxi deployments to over 15 cities by the end of this year.
- This move follows Uber's sale of its in-house autonomous vehicle unit in 2020 after internal struggles and a safety incident.
Uber Autonomous Solutions will provide software and services for operating autonomous vehicles and has been developing partnerships for years. The initiative follows Uber's earlier autonomous vehicle development struggles and aims to fill revenue gaps caused by shifts in the transportation market.
Read the original ↗
Google’s Cloud AI lead on the three frontiers of model capability
Understanding the challenges in AI model capabilities is crucial for businesses and developers to effectively implement artificial intelligence in diverse applications and industries.
- Michael Gerstenhaber, a product VP at Google Cloud, discusses how AI models face three frontiers: raw intelligence, response time, and scalability for cost-effective deployment.
- Gerstenhaber leads Vertex AI, Google's platform designed for customers to deploy enterprise AI solutions and access advanced model capabilities.
- The integration of Google's resources, including data centers and APIs, supports the development and deployment of agentic AI applications across various sectors.
Google Cloud's Vertex AI provides a platform for companies to build and manage AI applications, highlighting the need for ongoing infrastructure development and cost-efficiency in AI model deployment. The conversation with Gerstenhaber notes that agentic AI is still developing, with infrastructure and auditing patterns yet to be established. Observers should watch for advancements in model capabilities and their practical applications in business as the technology matures.
Read the original ↗
Anthropic accuses Chinese AI labs of mining Claude as US debates AI chip exports
Allegations of stolen AI technology can potentially heighten international tensions and impact U.S. policy decisions on AI chip exports to China, influencing global competition and security.
- Anthropic has accused Chinese AI labs DeepSeek, Moonshot, and MiniMax of using 24,000 fake accounts to distill its Claude AI capabilities.
- The labs reportedly generated over 16 million exchanges with Claude, targeting advanced features such as agentic reasoning and tool use.
- The accusations arise during debates over U.S. export controls on AI chips, which aim to limit China's AI development capacity.
- Anthropic claims that the scale of the alleged distillation requires access to advanced AI chips, reinforcing the need for stricter export regulations.
Anthropic is a prominent AI company known for its Claude model. The recent allegations highlight ongoing concerns regarding the unauthorized extraction of AI technology, which can undermine U.S. advancements. Future developments will likely focus on the implications for export controls and international security measures in AI technology.
Read the original ↗
OpenAI calls in the consultants for its enterprise push
OpenAI's new consulting partnerships aim to enhance enterprise adoption of its AI technologies, addressing a slow uptake in the industry and focusing on transforming business strategies.
- OpenAI is partnering with four consulting firms, including McKinsey and Accenture, to promote its enterprise AI platform, OpenAI Frontier.
- The initiative, termed 'Frontier Alliances', focuses on aligning AI with enterprise strategies and processes to drive effective adoption.
- OpenAI's Forward Deployed Engineering team will assist these consulting firms in integrating the Frontier platform into clients' technology stacks.
OpenAI Frontier, launched in early February 2026, is a no-code software that enables users to create and manage AI agents. The company's push for enterprise solutions comes as adoption of AI technology has been relatively slow due to concerns about return on investment, prompting partnerships with consulting firms to facilitate transformation in businesses.
Read the original ↗
How to Hide Google’s AI Overviews From Your Search Results
Users can avoid Google's AI-generated summaries in search results, which may enhance their search experience and access to unfiltered information.
- Typing "–ai" at the end of a Google search query removes AI-generated Overviews from the results.
- The trick works on computer browsers, with limited success on mobile apps like Safari and Chrome on iOS.
- Google allows a switch to Classic Search results that provide a mix of website links and short videos.
- Alternative search engines, such as DuckDuckGo and Brave, offer settings to toggle AI summaries on and off.
Google introduced AI Overviews in 2024, which have faced criticism for inaccuracies. The summaries are designed to enhance search utility, yet some users prefer traditional search formats without AI intervention.
Read the original ↗
The human work behind humanoid robots is being hidden
Understanding the hidden human labor behind humanoid robots could inform conversations about ethics and accountability in AI development, particularly in terms of transparency and user trust.
- Human contributions to the development of humanoid robots are often overlooked and not publicly acknowledged.
- The article discusses various AI terms that gained prominence in 2025, including 'superintelligence', and the investment that tech companies are making in AI.
- Concerns are rising about 'chatbot psychosis', where prolonged interactions with chatbots may lead to delusions or worsen mental health conditions.
Humanoid robots are designed to replicate human appearance and behavior, and their development has gained momentum in recent years due to advancements in AI. The increasing reliance on AI technologies raises important questions about transparency and the processes behind their creation. Future discussions are expected to focus on the ethical implications of AI technologies and their societal impact.
Read the original ↗
Guide Labs debuts a new kind of interpretable LLM
This new model addresses the challenge of understanding decisions made by AI systems, which is critical for ethical and practical deployments across various industries.
- Guide Labs has open-sourced an 8-billion-parameter language model (LLM) named Steerling-8B, designed to improve interpretability of AI-generated outputs.
- The model features a new architecture that allows users to trace each output token back to its training data origins.
- CEO Julius Adebayo, who began this research during his PhD, asserts that making models interpretable is now an engineering challenge rather than a scientific one.
- Steerling-8B claims to achieve 90% of the performance of leading models while using less training data, allowing for broader accessibility and control over AI outputs.
Guide Labs is a San Francisco startup focused on developing interpretable AI models. This technology is significant as it enhances transparency, particularly in regulated sectors like finance and healthcare. Moving forward, the company aims to scale its model and offer API access to users.
Read the original ↗
New Microsoft gaming chief has “no tolerance for bad AI”
Microsoft's new gaming chief's strong stance on AI in game development could influence industry standards and practices amidst ongoing debates over AI's role in creativity.
- Asha Sharma has been promoted to head Microsoft's gaming division following Phil Spencer's departure, having previously led the CoreAI Product group.
- Sharma expressed a strong opposition to what she terms 'bad AI' in game development, emphasizing the importance of human creativity in games.
- The gaming industry has seen backlash against AI-generated elements, exemplified by one studio losing awards for their use of generative AI in backgrounds.
Asha Sharma succeeds Phil Spencer as head of Microsoft's gaming division, emphasizing a focus on quality and human creativity in gaming. Her appointment follows scrutiny over her lack of gaming experience and comes during a significant transition in Microsoft's gaming strategies.
Read the original ↗
Spotify rolls out AI-powered Prompted Playlists to the U.K. and other markets
The rollout of Spotify's AI-powered Prompted Playlists can significantly enhance user experience by simplifying playlist creation and tailoring music to individual preferences.
- Spotify is launching its AI-powered Prompted Playlist feature for Premium subscribers in the U.K., Ireland, Australia, and Sweden.
- This feature allows users to create custom playlists by describing their desired music in their own words, catering to moods or scenarios.
- Once a prompt is submitted, Spotify utilizes AI to generate a personalized playlist, considering the user's listening history and current music trends.
Spotify, a leading music streaming platform, is expanding its AI features to improve user interaction and create a more personalized experience. The Prompted Playlist feature complements Spotify's ongoing developments in AI and aims to streamline how users engage with music recommendations.
Read the original ↗
Particle’s AI news app listens to podcasts for interesting clips so you you don’t have to
The ability for Particle's AI news app to extract relevant podcast clips enhances news consumption by providing listeners with concise, targeted information alongside articles, reflecting a significant shift in media engagement.
- Particle, an AI news app created by former Twitter engineers, now features Podcast Clips to extract key moments from podcasts related to news stories.
- Users can listen to short audio clips or read transcripts as they browse related news in the app.
- The app plans to monetize through a subscription model, offering premium features for $2.99 a month or $29.99 a year.
- Particle has expanded its reach internationally, with 55% of its users outside the U.S., and India being its second-largest market.
Particle is an AI-driven news application that aggregates content from both podcasts and traditional news to enhance user engagement. The newly introduced Podcast Clips feature reflects the growing trend of turning to podcasts for news and commentary. As media consumption evolves, tools like Particle are increasingly relevant for accessing timely information.
Read the original ↗
AIs can generate near-verbatim copies of novels from training data
The ability of AI models to generate near-verbatim copies of novels raises significant concerns about copyright infringement, which could affect ongoing legal battles and future training practices in the AI industry.
- Recent studies reveal that leading AI models from companies like OpenAI and Google can reproduce large portions of copyrighted novels with high accuracy.
- Researchers successfully prompted AI models to generate texts from titles such as 'Harry Potter' and 'The Hunger Games', revealing substantial memorization capabilities.
- The findings complicate the AI industry’s defense strategy against copyright lawsuits, as it contradicts past claims about how models handle copyrighted material.
Large language models (LLMs) are AI systems trained on vast amounts of text data, which has led to questions about their handling of copyrighted works. Previous assertions from the AI industry contended that models learn patterns rather than store data verbatim, but emerging evidence suggests otherwise. This ongoing issue continues to influence legal frameworks surrounding AI and copyright.
Read the original ↗
Defense Secretary summons Anthropic’s Amodei over military use of Claude
The Pentagon's threat to label Anthropic a "supply chain risk" could significantly impact the company's existing contracts and partnerships due to heightened scrutiny over military technology usage.
- Defense Secretary Pete Hegseth is meeting with Anthropic CEO Dario Amodei to address military applications of the AI system Claude.
- The Pentagon has threatened to designate Anthropic as a "supply chain risk" following the firm's refusal to allow its technology for domestic surveillance and autonomous weapons.
- Anthropic signed a $200 million contract with the Department of Defense last summer, and Claude was utilized in a special operations raid in January.
Anthropic is an AI firm known for developing Claude, an advanced AI system. The company's refusal to permit military surveillance and autonomous weapon applications has created tensions with the Pentagon. The outcome of this meeting may affect Anthropic's operational status and future contracts with the military.
Read the original ↗
How AI agents could destroy the economy
The potential economic consequences of integrating AI agents are highlighted as risks of rising unemployment and market value decline, which could have long-lasting impacts on the global economy.
- Citrini Research published a scenario on potential mass economic destruction caused by agentic AI over the next two years.
- The report predicts that unemployment could double and the stock market's total value could fall by more than a third.
- It describes a negative feedback loop where companies invest more in AI, leading to increased layoffs and reduced consumer spending.
Citrini Research is an analyst group exploring the impact of AI on the economy. Their report presents a hypothetical future where AI integration disrupts traditional employment and business models. As companies increasingly adopt AI technology, the effects on the workforce and economy warrant further investigation.
Read the original ↗
Data Storytelling using AI: 5 Techniques to Present AI-Generated In...
Effective data storytelling is essential for translating AI-generated insights into actionable business strategies, shaping decision-making processes in companies.
- AI can generate insights rapidly, but understanding their implications is crucial for business impact.
- Professionals must focus on what insights mean for the business, not just the AI processes used to obtain them.
- Five techniques are recommended for effective data storytelling, including starting with the business question and establishing the credibility of insights.
Data storytelling involves communicating insights derived from data in a way that resonates with business goals. With increasing reliance on AI for data analysis, clear communication of these insights becomes vital for decision-making. The skills required to translate data analysis into effective business strategies are now more crucial than ever.
Read the original ↗
AI’s Math Tricks Don’t Work for Scientific Computing
Understanding the differences in number formats for AI and scientific computing is crucial for optimizing performance and energy efficiency in simulations and computations across various fields.
- Low-precision number formats suitable for AI are not effective for scientific computing simulations.
- Laslo Hunhold, an engineer at Openchip, is developing a number format tailored for scientific computing called takum.
- Scientific computing requires high dynamic range and accuracy, which differs significantly from the needs of AI.
The growing demand for energy-efficient computing has led to the development of various number formats. AI often uses lower precision formats to save energy, whereas scientific computing requires high precision. The introduction of tailored formats like takum aims to bridge this gap in performance demands.
Read the original ↗
AI for Cybersecurity: Promise, Practice, and Pitfalls
The integration of AI in cybersecurity can enhance defense mechanisms but also introduces new vulnerabilities that organizations must address.
- AI is transforming cybersecurity by improving threat detection and enabling automated responses to attacks.
- While AI offers advancements, it also poses risks such as exploitation, manipulation, and bias in AI systems.
- The session will cover real-world applications of AI in cybersecurity, including anomaly detection and predictive modeling.
Artificial intelligence is increasingly utilized in cybersecurity to combat advanced threats, but its implementation presents various challenges. As organizations adopt AI technologies, understanding the associated risks becomes crucial. Upcoming discussions will focus on responsible AI use in this sector.
Read the original ↗
Hitachi bets on industrial expertise to win the physical AI race
Hitachi's focus on leveraging its industrial expertise signals a strategic move to compete effectively in the growing physical AI market, which could influence industry standards.
- Hitachi is emphasizing its industrial expertise to enhance its position in the physical AI sector.
- The company aims to differentiate itself by applying its existing knowledge to develop robust AI solutions.
Hitachi is a multinational company known for its diverse industrial products and technology solutions. The physical AI race involves developing artificial intelligence technologies that interact with and influence the physical world, which is becoming increasingly competitive.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Large-Scale Terminal Agentic Trajectory Generation from Dockerized Environments
The development of TerminalTraj represents a significant advancement in training agentic models, enabling improved performance across diverse tasks and enhancing the capability of automated systems.
- TerminalTraj is a scalable pipeline that generates high-quality terminal trajectories for training agentic models by creating Docker-aligned task instances.
- The pipeline addresses challenges of executability and verifiability by filtering repositories and synthesizing agent trajectories with executable validation code.
- Over 50,733 verified terminal trajectories were generated across eight domains, improving model performance by up to 20% on the TerminalBench benchmark.
TerminalTraj aims to facilitate the training of agentic models, which rely on realistic interactions captured through terminal trajectories. Traditionally, generating such data at scale has posed significant challenges due to diverse and complex environments. The advancements made here may set a precedent for future technologies in automated task execution.
Read the original ↗
Taalas is replacing programmable GPUs with hardwired AI chips to achieve...
The shift from programmable GPUs to hardwired AI chips may significantly enhance processing speeds for AI applications, impacting performance across various industries.
- Taalas is replacing programmable GPUs with hardwired AI chips.
- This change aims to achieve 17,000 tokens per second for ubiquitous inference, or consistent AI processing across various devices.
Taalas is a company focused on advancing AI technology. Traditionally, GPUs, or graphical processing units, have been used for AI processing, but this transition represents a shift towards more specialized hardware. Future developments will likely focus on the performance measures and applications of these hardwired chips.
Based on the headline only — the full article could not be fetched.
Read the original ↗