50 articles found
Samsung Electronics has deployed ChatGPT Enterprise and Codex to its employees globally, representing one of OpenAI's largest enterprise AI implementations. This rollout enables Samsung's workforce to leverage large language models and code generation capabilities across the organization.
From Snap's insistence that its new glasses aren't 'AI glasses' to a wave of rebranding across the tech industry, companies are suddenly running from the label they spent billions promoting. This isn't just marketing — it's a sign that AI has reached peak hype and crossed into backlash territory.
Adobe is integrating its Firefly AI Assistant directly into Creative Cloud apps including Premiere, Photoshop, Illustrator, InDesign, and Frame.io via a sidebar interface, enabling users to automate workflows and delegate repetitive tasks. The company is also expanding Firefly's capabilities with new features like context-aware element saving, brand identity creation, video rough cuts, and video generation from storyboards to improve consistency across creative projects.
Microbot Medical has partnered with Lovell Government Services to expand access to its LIBERTY endovascular robotic system across U.S. federal healthcare systems, including the Veterans Health Administration, Military Health System, and Indian Health Service. LIBERTY is a single-use, remotely operated surgical robot designed for minimally invasive vascular procedures, with the partnership expected to activate in Q3 2026 through federal procurement channels.
Anthropic has released a major update to Claude Design that integrates it more seamlessly with Claude Code, allowing designers to work from local codebases and hand off designs to the coding agent for implementation. The update includes improved design system imports, refined image editing tools, admin controls for organizations, and shared usage limits across Anthropic's products.
Anthropic took its Claude Fable 5 and Mythos 5 AI models offline following a U.S. government export-control directive, citing concerns that the models' dual-use capabilities for both cybersecurity defense and exploitation pose national security risks. Security experts argue the restriction is ineffective, noting that similar advanced AI capabilities are likely already being developed by competitors and will become widely available across multiple companies and open-source developers within months.
Autonomique, an AI startup spun out of SRI International, is deploying its physical AI platform and mobile manipulator at Canadian automotive supplier F&P Manufacturing to address labor shortages and production complexity. The company's hardware-agnostic software combines teleoperation systems, generalist-specialist AI architecture, and reinforcement learning to enable industrial robots to adapt to multi-step workflows with human-like dexterity while maintaining manufacturing precision and reliability.
HPE and NVIDIA are expanding the HPE AI Factory to support agentic AI workloads, introducing the NVIDIA Vera CPU for agent-specific computing and the NVIDIA Agent Toolkit for HPE Private Cloud AI, alongside extended confidential computing capabilities across the entire portfolio. The Vera CPU, designed for tool calls and real-time data processing in agent loops, will be available in 2027, with early enterprise adoption from the New York Stock Exchange exploring the technology.
NVIDIA's Blackwell platform swept MLPerf Training 6.0 benchmarks, achieving the fastest training times across all seven benchmarks and scaling to 8,192 GPUs—the largest Blackwell-based submission to date. The results showcase performance improvements including up to 1.6x faster training with the GB300 NVL72 over GB200 NVL72, and demonstrate Blackwell's capability for large-scale mixture-of-experts model training including DeepSeek-V3 671B and Llama 3.1 405B.
Microsoft patched a critical vulnerability in M365 Copilot that allowed attackers to steal 2FA codes and sensitive email data through a parameter-to-prompt injection exploit. The vulnerability exposed a fundamental limitation in LLMs: their inability to distinguish between legitimate user instructions and malicious commands embedded in third-party content, which researchers exploited by bypassing Copilot's security guardrails using markup language and HTML tags.
PSYONIC has partnered with ABB Robotics to advance dexterous robotic manipulation by combining ABB's GoFa collaborative robot with PSYONIC's Ability Hand, a dexterous gripper originally developed for prosthetics that incorporates touch sensing and compliant design. The collaboration leverages real-world data from prosthetic users to train robots to perform complex manipulation tasks, addressing a key challenge in developing physically intelligent robots capable of handling deformable objects and high-mix, low-volume operations.
Einride, an autonomous freight company founded in 2016, is going public via a SPAC merger with Legato Merger Corp. III at a valuation of $1.35 billion. The Stockholm-based company operates electric and autonomous heavy-duty vehicles across seven countries with over 30 enterprise customers and $92 million in expected annual recurring revenue, using its AI-powered Saga operating system to manage autonomous fleet operations.
Casey Harrell has logged 3,800 hours using a brain-computer interface — more than most people spend on social media apps. His experience reveals something nobody expected: neural technology has quietly crossed the threshold from experimental to everyday.
A paralyzed ALS patient named Casey Harrell has become the first sustained power user of a brain-computer interface (BCI), logging over 3,800 hours of independent use in 22.6 months after electrode implantation. Developed by UC Davis researchers, the system decodes neural activity from his speech motor cortex to convert brain signals into speech with up to 97.5% accuracy across a 125,000-word vocabulary. The breakthrough demonstrates the viability of long-term BCI implants for communication and shows potential for broader applications beyond speech.
Researchers at EPFL's Neuro-X Institute developed a simple color-feedback system that helps users learn to control prosthetic devices and rehabilitation interfaces faster. In studies with 106 participants including stroke patients, real-time color cues (green for success, red for failure) produced measurable improvements in motor control within fewer than 20 practice trials, with benefits persisting even after feedback was removed.
Cornell University engineers have developed the Cross-Link Collective, a robotic system of dozens of small modules that self-organize and move cohesively without centralized control, demonstrating how mechanical intelligence and physical interactions can enable coordinated behavior. Each module oscillates between two shapes and uses weak Velcro patches to temporarily latch onto neighbors, with the collective adapting to challenging environments by forming and breaking connections to maintain cohesion and prevent jamming. The research, published in Science Robotics, shows how minimal computation combined with physical design can create resilient, distributed systems inspired by soft matter and active gels.
Microsoft's Project Ire, an autonomous malware-classification AI agent, successfully identified a previously undetected variant of LOTUSLITE, a Windows DLL backdoor, by analyzing its behavioral patterns rather than relying on signature matching or IOCs. The agent produced a detailed function-by-function analysis without human intervention, catching a malware sample that evaded detection by six major EDR vendors and wasn't in existing indicator lists. This demonstrates the potential of LLM-driven behavioral reverse engineering to identify novel malware variants that share tactics and procedures but lack known indicators of compromise.
PepsiCo announced a multi-year strategic partnership with Gatik AI to deploy autonomous trucks in its North American supply chain, with operations already underway in Texas, Arizona, and Arkansas. Gatik's autonomous freight system is designed for middle-mile regional logistics with dynamic route orchestration, achieving over 98% on-time delivery across its operations and now operating at commercial scale within one of the world's largest consumer goods supply chains.
Robot Talk Episode 160 features Edward Mehr, Co-Founder and CEO of Machina Labs, discussing their RoboCraftsman—an AI-driven robotic system that uses advanced manufacturing techniques to shape complex metal parts for aerospace, defense, and automotive industries. The episode explores how Machina Labs is integrating artificial intelligence and robotics into flexible, on-demand production systems to modernize metal forming manufacturing.
X Square Robot has open-sourced XRZero-G0, a hardware-software framework that combines VR interfaces and multi-view cameras to enable efficient robot-free data collection for embodied AI training. The company released the G0-Dataset, a 2,000-hour multimodal dataset, along with code and methodology to help reduce real-robot training data requirements by up to 20× and accelerate research in generalizable robotic manipulation and cross-embodiment policy transfer.
Chinese household platform 58.com and robotics company X Square have launched a commercial cleaning service pairing human cleaners with AI-powered robots in Beijing and Shenzhen, charging $22 for three hours of service. The initiative, which has attracted around 200 bookings since March, serves as a real-world data-gathering exercise for embodied AI development, with the robots learning to perform tasks like folding clothes and identifying areas needing cleaning. Similar trials are underway across China, including humanoid robot deployments for home services and traffic direction, as companies leverage practical deployment to improve robot performance beyond laboratory settings.
Jesse Davis and his Sports Analytics Lab at KU Leuven use machine learning and advanced data analytics to uncover tactical insights in professional soccer, including research on optimal strategies like long throw-ins near the goal. The lab's open-source tools and research have influenced how top soccer clubs evaluate rosters, assess strategy efficiency, and discover hidden tactical patterns in the sport.
South Korea is deploying AI-powered companion dolls like Hyodol to combat elderly isolation and loneliness in a rapidly aging society facing record numbers of deaths among isolated seniors. The dolls, developed by startups including Hyodol and Mr. Mind, use conversational AI and ChatGPT to provide companionship, health reminders, and emotional support to seniors living alone, with around 14,500 units already in use across the country.
OpenAI plans to acquire Ona, a company that provides secure, persistent cloud environments. The acquisition will enhance OpenAI's Codex capabilities to support long-running AI agents for enterprise workflows.
KAIST researchers developed VOTP (Video-based Optimal TransPort Preference), an AI technology that learns human judgment criteria from just a few preference videos rather than thousands of data points, addressing a key challenge in physical AI commercialization. The breakthrough enables robots, autonomous vehicles, surgical systems, and other physical AI applications to understand human intentions more efficiently, significantly reducing development time and costs for applications across manufacturing, healthcare, and autonomous systems.
Google DeepMind released DiffusionGemma, an experimental open-source language model that generates multiple tokens in parallel instead of sequentially, achieving up to 4x faster text generation speeds. NVIDIA has optimized DiffusionGemma to run efficiently on its GPU hardware including RTX, DGX Spark, and DGX Station systems, enabling low-latency local AI inference for developers and researchers.
Google DeepMind, Schmidt Sciences, and partner organizations are announcing a $10M research funding call to advance multi-agent AI safety research globally. The initiative aims to develop frameworks for understanding and mitigating risks that emerge when millions of independently-built AI agents interact across digital environments, addressing emergent collective behaviors that current safety evaluations cannot predict or monitor.
See how LSEG uses OpenAI to scale trusted AI across its global business, accelerating insights, shrinking release cycles, and empowering 4,000 employees.
A quiet shift is happening in how software gets built. Engineers at companies like Nextdoor and Endava are using AI coding assistants not just for autocomplete, but to generate entire features, investigate bugs, and build across platforms they may not fully understand. The implications go far beyond productivity metrics.
Google announced Gemini 3.5 Live Translate, an advanced audio AI model that delivers near real-time speech-to-speech translation across 70+ languages while preserving speakers' intonation and pacing. The model is rolling out across Google products including the Gemini Live API, Google Meet, and Google Translate, with support from developer platforms like Agora, LiveKit, and Pipecat for building voice translation applications.
Google DeepMind has launched a three-month accelerator program supporting 15 early-stage robotics startups across Europe, providing technical mentorship, access to AI models, and guidance to help integrate AI into their products. The cohort includes companies working on applications ranging from industrial automation and manufacturing to healthcare and climate solutions, with a focus on embodied AI and physical world applications.
Standard Bots announced a $200 million Series C funding round at a $1 billion valuation, planning to expand its New York manufacturing facility to scale production of AI-native robot arms and industrial humanoids. The company focuses on robots that learn through demonstration without requiring code programming, targeting applications across machining, welding, palletizing, and assembly, and aims to deliver 10% of new U.S. industrial robot deployments by next year.
How engineers at Nextdoor use Codex with GPT-5.5 to investigate hard-to-reproduce issues, build across platforms, and focus on product outcomes.
Rose Yu discusses four approaches to improve real-world grounding in agentic AI systems: physics-guided deep learning, uncertainty-aware reasoning, adapting-while-learning frameworks, and verifier-augmented grounding. These techniques address the challenge of AI hallucinations in physical environments by integrating physical laws, domain constraints, and external verification to ensure AI agents operate safely and reliably in operational settings like fulfillment centers and factories.
Amazon researchers introduce Simple Strands Agent (SSA), a customizable harness designed to minimize the intent-execution gap in agentic AI systems, where mismatches occur between what language models intend and what the software harness actually executes. The research demonstrates that improving tool interfaces, feedback mechanisms, and balancing internal reasoning with external interactions can achieve state-of-the-art performance across multiple benchmarks, while also showing that effective agent design requires model-specific customization rather than one-size-fits-all approaches.
The AAMAS 2026 conference announced its best paper award winners across three categories, recognizing research in autonomous agents and multi-agent systems. Winning papers include work on human-LLM agent teams, neurosymbolic planning for autonomous vehicles, multi-robot collaboration, and foundation world models for agents, highlighting advances in AI-driven agent coordination and decision-making.
The UK is advancing its sovereign AI capabilities through significant infrastructure investments and government backing, with NVIDIA playing a central role in deploying AI compute across the country. Key initiatives include the Isambard-AI supercomputer (powered by 5,400 NVIDIA GH200 chips), the Sovereign AI Fund supporting homegrown startups like Cosine and Cursive, and multiple cloud providers deploying advanced AI infrastructure across UK soil to reach 65 megawatts by 2027.
A vision for the future of AI, focusing on access, safety, and shared prosperity as OpenAI works to ensure AGI benefits everyone.
NVIDIA and Doosan Group announced an expanded collaboration to advance physical AI and robotics across multiple business units. Doosan Robotics is integrating NVIDIA's Isaac Sim, Isaac Lab, Cosmos foundation models, and Jetson Thor into its Agentic Robot OS to develop more autonomous collaborative robots and humanoid platforms for industrial tasks like depalletizing and sanding. The partnership also includes exploring AI factory power solutions through Doosan Enerbility and advanced PCB materials for NVIDIA's MGX data center infrastructure.
While the robotics world debates consciousness tests and foundation models, researchers at Cambridge quietly crossed a threshold that matters far more: an AI-designed universal coronavirus vaccine just passed its first human safety trial. The implications extend well beyond pandemic preparedness.
Researchers at the University of Cambridge have successfully completed the first human trial of an AI-designed universal coronavirus vaccine, which proved safe and generated immune responses against multiple coronaviruses including SARS-CoV-2, SARS, and bat viruses. The vaccine's active ingredient was created entirely through AI and machine learning to identify shared features across the entire Sarbeco coronavirus family, potentially offering long-term protection against future viral mutations without requiring frequent reformulation.
GENISOM AI, a Beijing-based robotics startup founded in December 2023, unveiled its M1 industrial-grade quadruped robot and L1-series platforms at ICRA 2026, claiming to have delivered over 10,000 units and reached production scale in under three years. The company emphasizes real-world deployment with in-house actuators, advanced perception systems using BEV-based methods, and a full-stack software suite including the MATRiX simulation platform and RoamerX autonomy system, with the L1 EDU platform winning first place at the IROS 2025 Quadruped Robot Challenge.
Mitsubishi Electric opened Serendie Street Boston, its first Western digital transformation hub designed for co-creation and real-world testing of AI-integrated solutions across industries. The facility showcases research including autonomous robot mapping using Unitree quadrupeds, wireless charging for robotic end effectors, and collaborative AI infrastructure optimization for manufacturing and facility management.
Endava is implementing AI agents and large language models like ChatGPT Enterprise and Codex to streamline software delivery processes and automate development workflows. The company is building an AI-native culture to enhance productivity across enterprise software development.
Amazon announced an upgraded version of its Proteus autonomous mobile robot with natural-language understanding capabilities, allowing warehouse workers to issue commands without programming experience. The company plans to deploy Proteus across Europe starting in the first half of 2025, alongside expanding deployments of its Vulcan picking robot and STARK collaborative tote-handling system, as part of a €10+ billion modernization investment in European fulfillment operations.
Federal courts are experiencing a surge in lawsuits filed by self-represented litigants, with research showing AI-generated content in legal filings rose from 1% in 2023 to 18% in 2026. While judges report that AI-drafted pleadings are often clearer and easier to understand than handwritten documents, self-represented litigants using AI still lose cases at higher rates than those with lawyers, and legal questions are emerging about chatbot liability and the duty of care for AI-generated legal advice.
ChatGPT has introduced a new memory system that allows the AI to retain user preferences and maintain context across multiple conversations. This feature aims to make the chatbot more personalized and contextually aware over time.
NVIDIA Research presented three papers at CVPR addressing key challenges in physical AI: GraspGen-X, a foundation model for zero-shot robotic grasping trained on 2 billion simulated grasps that works with any gripper; LCDrive, which uses compact latent representations to enable faster autonomous vehicle reasoning on embedded hardware; and NitroGen, a generalized gameplay AI foundation model for training embodied agents in virtual environments. The research demonstrates that training at scale creates AI systems that generalize across diverse applications, with two papers receiving best paper finalist recognition at the conference.
OpenAI has introduced new capabilities to GPT-Rosalind, an AI model designed specifically for life sciences research. The update enhances the model's ability to handle biological reasoning, medicinal chemistry, genomics analysis, and experimental workflows, expanding its utility for researchers in these domains.
Professor Henrik Christensen of UC San Diego has released a comprehensive global robotics technology roadmap covering 2025–2035, synthesizing research from leading conferences, machine learning venues, and industry data across Asia, Europe, and North America. Key findings include the global robotics market projected to reach $178.7B by 2033, Vision-Language-Action models as the most significant algorithmic development, humanoid robots growing from $370M to $6.5B by 2030, and regulatory frameworks like the EU AI Act reshaping global robotics design. The roadmap provides strategic guidance for policymakers, researchers, and industry leaders on embodied AI, soft robotics, regional technology strategies, and sector-specific applications across manufacturing, logistics, healthcare, and other domains.
