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
Google announces Gemini 3.6 Flash and cybersecurity AI, teases 3.5 Pro and Gemini 4
These updates signal Google's continued advancements in AI capabilities, particularly in efficiency and cybersecurity, critical for both developers and enterprise users navigating rising AI costs.
- Google has released Gemini 3.6 Flash, upgrading from the previously released 3.5 version, enhancing coding abilities and using 17% fewer tokens.
- The new Gemini 3.5 Flash Lite targets efficient scaling of AI systems, performing at 350 tokens per second with competitive pricing.
- Gemini 3.5 Flash Cyber, aimed at cybersecurity applications, will initially launch as a limited pilot for trusted partners and governments.
- Google is currently testing Gemini 3.5 Pro, previously delayed, which aims to compete with leading models like GPT 5.6 and Claude Fable/Sonnet 5.
- Pre-training for Gemini 4 has begun, signaling the next phase of Google's AI model development without a specified release timeline.
Google is a major player in AI development, recently announcing several new AI model releases. The Gemini series focuses on enhancing efficiency and capabilities in various fields, particularly in coding and cybersecurity. Future developments will continue to evolve as Google tests new models and expands capabilities.
Read the original ↗
Anthropic’s $1.5B copyright settlement approved; only 350 authors opted out
The approval of Anthropic's $1.5 billion copyright settlement underscores significant implications for the rights of authors against AI training practices, marking a landmark shift in copyright law.
- A judge has approved a $1.5 billion settlement between Anthropic and authors, the largest copyright class-action ever certified.
- Only 350 authors opted out of the settlement, while about 91 percent of authors involved filed claims and supported the settlement.
- The judge reduced lawyers' fees from 20 percent to less than 7 percent of the settlement fund, amounting to about $101 million.
- Both the lead plaintiffs and most authors indicated satisfaction with the settlement's provisions, despite some objections regarding payouts.
- The court emphasized the importance of timely notices, denying most late opt-out requests to maintain the settlement's finality.
Anthropic is an AI research company that had faced allegations of copyright infringement for training its AI on copyrighted texts. The settlement followed a ruling that defined the legal boundaries of fair use in AI training while addressing author compensation. Future developments may involve further redistribution of settlement funds if applicable claims allow.
Read the original ↗
Validating Distributed LLM Serving Benchmarks with NVIDIA srt-slurm, SLURM Recipes, Parameter Sweeps, and Pareto Analysis
Understanding NVIDIA's srt-slurm framework enables developers to effectively validate and prepare distributed LLM-serving benchmarks, optimizing performance before deployment on GPU clusters.
- The article discusses the use of NVIDIA's srt-slurm framework for creating reproducible benchmark workflows for distributed large language model (LLM) serving.
- It describes setting up the project in Google Colab to simulate and validate configurations before actual deployment on a GPU cluster.
- Key methods include using parameter sweeps and analyzing benchmark results through a throughput-versus-latency Pareto analysis.
NVIDIA's srt-slurm is a framework designed for managing and optimizing resource allocation for distributed computing tasks, particularly in AI applications. Developers previously faced challenges in validating performance before deployment, prompting the exploration of tools like srt-slurm for improved efficiency. Future developments will focus on enhancing the workflow for transitions to actual GPU cluster environments.
Read the original ↗
Gritt Launches with $32.4M to Bring Physical AI to Infrastructure Construction
The development of AI-powered robots in construction could significantly address labor shortages and improve efficiency in major infrastructure projects amid rising demand for construction services.
- Gritt has raised $32.4 million in funding to create AI-powered robotic systems for construction tasks.
- The company is initially deploying its technology in large-scale solar projects to enhance efficiency and reduce labor needs.
- Gritt's machines are designed to connect with existing construction equipment, simplifying integration and maintenance.
- The technology aims to automate repetitive tasks and provide real-time data for better project management decisions.
Gritt is a startup focused on developing robotics for the construction sector, facing challenges from labor shortages and productivity issues. Its technology is suitable for unpredictable construction environments, promising to transform how projects are executed. Future expansions may include data centers and broader infrastructure work.
Read the original ↗
Humanoid Raises $152 Million to Scale Industrial Humanoid Robots
Humanoid's substantial funding could accelerate the deployment of industrial humanoid robots, potentially transforming how tasks are performed in logistics and manufacturing environments.
- London-based robotics company Humanoid has raised $152 million in Series A funding, with a post-money valuation of $1.35 billion.
- The funding aims to develop Humanoid's next generation of industrial robots and expand customer deployments, including a significant deal with Schaeffler.
- Humanoid has already executed nine industrial deployments and has 34,000 robot pre-orders, with a pipeline valued at approximately $2.4 billion.
Humanoid develops general-purpose robots for industrial applications, focusing on logistics and manufacturing. Previously, the company emphasized workplace efficiency with robots designed for existing infrastructure. Looking ahead, the next phase involves scaling production and addressing practical operational challenges.
Read the original ↗
Vijay Vijayasankar, Global Agentic AI Officer at Genpact – Interview Series
This interview with Genpact's AI leader highlights critical insights about successfully integrating AI into business operations, particularly in overcoming enterprise debt that hinders value realization.
- Vijay Vijayasankar is the Global Agentic AI Officer at Genpact, focusing on the productization of Agentic AI solutions to enhance enterprise value creation.
- Genpact emphasizes transitioning from traditional manpower-based models to outcome-based frameworks, particularly in areas such as finance and customer care.
- A recent report estimates that nearly $18 trillion in potential value is trapped due to unresolved enterprise debt, which includes inefficient processes and fragmented data.
- Vijayasankar argues that effective AI transformation requires organizations to address foundational issues like data quality and process efficiency, rather than just technology deployment.
- He emphasizes that scaling AI solutions involves leadership accountability and structural change, rather than merely implementing new software.
Genpact is a technology solutions company that helps large enterprises modernize operations through AI, data, and digital transformation. Historically, companies have struggled to realize AI's full potential due to underlying operational inefficiencies and poor data governance. Future developments to watch include Genpact's evolving strategies for integrating AI within various business functions.
Read the original ↗
WEKA Launches NeuralMesh 6 and WEKApod 3 as AI Infrastructure Shifts Toward Inference
The advancements in WEKA’s infrastructure could enhance AI model efficiency, enabling better utilization of GPUs and improving performance for various applications.
- WEKA launched NeuralMesh 6 and WEKApod 3 to improve AI infrastructure for inference, addressing challenges of keeping GPUs productive.
- NeuralMesh 6 offers a unified platform for managing data storage, memory, and access without creating multiple copies of the same data.
- WEKApod 3 features a new design for enhanced performance and capacity, suitable for demanding GPU workloads and AI applications.
WEKA is focusing on AI infrastructure, moving from traditional data storage methods to a more integrated approach that supports high-performance computing needs. This reflects ongoing shifts in AI technology as the industry demands more efficient methods for managing AI workloads, particularly in response to large-scale inference challenges.
Read the original ↗.jpg)
Nvidia Wants to Own Every Chip Inside AI Data Centers
Nvidia's strategy to dominate AI data center chip supply could significantly alter the competitive landscape in the semiconductor industry, impacting performance and prices.
- Nvidia has unveiled its Vera Rubin chip system, combining GPUs and CPUs to enhance AI infrastructure performance.
- The Vera system features 36 CPUs for every 72 GPUs, designed for improved efficiency and reduced installation time.
- Nvidia claims Vera Rubin processes 10 times more tokens per watt compared to previous models, targeting market share from AMD and Intel.
- The new liquid-cooled design minimizes energy use and streamlines connectivity, supporting easier system integration.
- Nvidia is positioned to ship Vera Rubin in the second half of 2026, with early adoption by companies like OpenAI and Microsoft.
Nvidia is a leading supplier of graphics processing units (GPUs) and is expanding into the central processing unit (CPU) market to meet the growing demand for AI capabilities. The company has previously faced challenges with its older chip models, necessitating design adjustments. Future developments to watch include competitor responses and performance benchmarks from AMD's upcoming product announcements.
Read the original ↗
New Research Reveals How Traveling Brain Waves Allow Us to Predict, Reconstruct, and Perceive the World
This research sheds light on how brain waves contribute to our ability to perceive and understand the world, potentially influencing future studies on neurological function and cognition.
- Scientists at the Salk Institute found that traveling brain waves in the visual cortex enhance prediction and perception capabilities.
- The research suggests that similar brain waves may exist in other sensory systems beyond visual processing.
- The team proposes that these waves help create internal representations of the outside world by modifying synaptic connections based on experiences.
The Salk Institute focuses on neuroscience research, particularly brain function. Previous studies had established the importance of neural traveling waves in behavior and attention. This research provides a deeper understanding of how these waves can shape our perception of reality.
Read the original ↗
Artificial Intelligence (479)
Understanding how to run enhanced coding models locally can empower developers to work more efficiently and leverage advanced AI tools without relying on cloud resources.
- The article discusses running the Mythos enhanced coding model locally using llama.cpp and Raspberry Pi.
- It highlights techniques for developers to implement advanced coding assistance directly on their systems.
Artificial intelligence in coding aims to improve how software is developed by providing intelligent assistance and automating tasks. Local implementations can help reduce reliance on cloud computing and increase privacy.
Based on the headline only — the full article could not be fetched.
Read the original ↗
US threatens sanctions against Chinese AI models over IP theft
The U.S. government's potential sanctions against Chinese AI models highlight growing tensions over intellectual property theft and could impact the future landscape of AI innovation.
- Treasury Secretary Scott Bessent announced the U.S. is examining Chinese open source AI models for signs of intellectual property (IP) theft.
- Sanctions may target AI models if evidence of IP theft is found, expanding efforts to limit China's AI advancements.
- Concerns have been raised that Chinese models like Moonshot AI's Kimi K3 threaten leading U.S. firms such as OpenAI and Anthropic.
The U.S. government is exploring measures to curb intellectual property theft linked to Chinese AI models, a continuation of actions aimed at maintaining competitiveness in AI technology. This follows restrictions on China’s access to sophisticated semiconductor technology and intensified scrutiny of AI development practices.
Read the original ↗
HOMIE: Human-object Centric Video Personalization via Multimodal Intelligent Enchancement
HOMIE represents a significant advance in human-object video personalization, potentially improving the accuracy of video generation for interactions involving diverse objects, which has implications for various multimedia applications.
- HOMIE is a framework designed for human-object centric video personalization that enhances interaction fidelity and reference alignment.
- The framework addresses limitations in existing methods by balancing subject fidelity with accurate interaction patterns between humans and objects.
- HOMIE integrates global multimodal guidance to align semantic features with video generation tokens, enhancing personalization capabilities.
- The framework has been validated through extensive experiments showing state-of-the-art performance across various human-object video personalization tasks.
Human-object centric video personalization (HOCVP) focuses on generating videos tailored to specific subjects interacting with objects. Previous methods have faced challenges in fidelity and reference understanding. Future developments in this area may further refine video personalization techniques.
Read the original ↗
Spectro Cloud Launches PaletteAI Inference Launchpad to Help Enterprises Control AI Token Costs
Enterprises face rising AI costs as they scale, making effective management of AI token consumption critical to controlling infrastructure expenses and operational efficiency.
- Spectro Cloud launched PaletteAI Inference Launchpad to help enterprises better manage AI inference costs, potentially reducing token costs by up to 70% based on various factors.
- The platform allows organizations to run suitable workloads on their own infrastructure instead of relying solely on external AI services, enhancing cost control and reducing latency.
- PaletteAI supports a hybrid model, enabling enterprises to utilize both local and external AI models according to task complexity and requirements.
Spectro Cloud develops AI infrastructure management solutions, facilitating the deployment and maintenance of AI models in various environments. The recent launch reflects a trend in enterprise AI toward controlling costs while implementing AI functionalities across operations. The company has also expanded its support for AMD infrastructure alongside its product launch.
Read the original ↗
Photon Queue Raises $4M to Bring Quantum Memory Hardware Into Real-World Systems
Photon Queue's $4 million funding will accelerate the practical application of quantum memory technology, potentially enhancing quantum computing and communication systems significantly.
- Photon Queue has raised $4 million in a seed funding round to advance quantum memory technology for real-world applications.
- The funding round was led by Playground Global and enables Photon Queue to expand product development and customer installations.
- Photon Queue's technology stores photons in free-space loops without requiring cryogenic cooling, making it suitable for various quantum computing architectures.
Photon Queue, established in 2024 as a spinout from the University of Illinois Urbana-Champaign, focuses on developing quantum memory solutions essential for synchronizing quantum operations. The company's recent collaborations with Sandia National Laboratories and the University of Maryland enhance its product testing and commercial readiness.
Read the original ↗
Music streamer Deezer says more than 50% of daily uploads are AI-generated
The rise of AI-generated music, now constituting over half of daily uploads on Deezer, highlights a significant shift in the music industry and necessitates new policies for artist rights and platform management.
- Deezer reports that more than 50% of daily uploads on its platform are now AI-generated tracks, with a peak of 90,000 such uploads recorded in June 2026.
- The number of AI-generated tracks uploaded daily on Deezer has increased significantly from 10,000 per day in January 2025 to 90,000 in June 2026.
- In response to the rise of AI music, Deezer will remove AI-generated tracks that lack streaming activity for six months or have fraudulent streams.
- Deezer has been actively combating fraud and dilution of artist payments from AI music since 2025 and has implemented a labeling system for AI-generated content.
Deezer is a music streaming service that began tracking AI-generated track uploads last year. With a growing number of platforms responding to the influx of AI music, Deezer's actions may influence broader industry standards and practices in managing AI-generated content.
Read the original ↗
OpenAI Says Codex and ChatGPT Work Hit 10 Million Users
OpenAI's growth to 10 million users for Codex and ChatGPT Work indicates significant interest in AI-driven task completion tools, highlighting competitive dynamics in the AI market.
- OpenAI reports that 10 million people are now using Codex and ChatGPT Work, a notable increase since launching ChatGPT Work on July 9, 2026.
- The combined user count roughly doubled in just two weeks, indicating rapid adoption of these AI products.
- The growth includes a significant percentage of non-developers, suggesting a shift in market demographics for AI tools.
- This surge in usage follows the release of OpenAI's GPT-5.6 model, which has attracted paying users.
- OpenAI's usage figures are less clear in terms of retention and activity definitions, which raises questions about long-term adoption.
OpenAI provides AI technologies including Codex, a coding agent, and ChatGPT Work, aimed at enhancing productivity for a broader user base. This growth occurs amid heightened competition from other AI companies, particularly Anthropic, as they launch similar products targeting enterprise customers.
Read the original ↗
Your AI Budget Is Writing Checks Your Cryptography Can’t Cash
Organizations face potential financial risks if they fail to prioritize post-quantum cryptography adoption, which is essential to protect sensitive data from future quantum computing threats.
- Leading organizations such as Moody's, Google, and Cloudflare warn that neglecting post-quantum cryptography (PQC) may result in severe financial risks for various sectors.
- The White House has mandated accelerated timelines for federal agencies to transition to PQC, with deadlines now imposed for the protection of critical systems.
- Research indicates that advancements in quantum computing could shorten the timeline for when current encryption methods will be vulnerable, reducing estimates of required quantum computing power for attacks.
Post-quantum cryptography (PQC) is a type of encryption designed to protect against potential threats from quantum computers. Recent years have seen rising concerns about the implications of quantum computing, prompting accelerated efforts from corporations and governments to transition to quantum-resistant protocols. Organizations are urged to prioritize their cryptographic migration strategies to mitigate risks from emerging technologies.
Read the original ↗
AI Is Alive. Your Risk Just Changed.
As AI systems begin to operate autonomously within enterprise environments, the nature and immediacy of cybersecurity risks change significantly, necessitating urgent attention to governance and control measures.
- AI is starting to act and make decisions autonomously within enterprise systems, significantly altering the risk landscape.
- Systems from companies like Anthropic, OpenAI, and Google DeepMind show AI's ability to operate at machine speed, compressing decision-making timelines.
- Three key risk pathways identified include autonomous system manipulation, persistent system compromise, and self-exfiltration beyond enterprise controls.
AI, previously a tool, is evolving into an autonomous actor in cybersecurity. This shift brings unique risks that traditional systems may not adequately address, highlighting the need for better safety frameworks.
Read the original ↗
The Possible Demise of a Federal Program Isn’t a Rejection of AI in Healthcare
The potential defunding of the WISeR program highlights the challenges of integrating AI in healthcare, emphasizing the need for human oversight despite AI's efficiency promises.
- The House Appropriations Committee voted to block funding for the WISeR program, which uses AI to expedite claims adjudication in Medicare.
- WISeR is currently active in six states, focusing on increasing efficiency in healthcare financing.
- Lawmakers raised concerns that WISeR's introduction of prior authorization increased administrative complexity in patient care.
- The principle of 'AI flags, clinician affirms' continues to guide the governance of AI applications in healthcare, regardless of WISeR's future.
AI is rapidly transforming healthcare by enhancing efficiency, particularly in administration and diagnostics. The WISeR program exemplifies the innovative use of AI but faces governance and funding challenges. Ongoing discussions regarding AI's integration into healthcare will center on balancing technology with human oversight.
Read the original ↗
The Army Is Burning Through Its AI Tokens
The Army's rapid depletion of AI tokens raises concerns about resource management and effectiveness in using generative AI technologies for military operations.
- Members of the Army's DEVCOM were informed of rapidly depleting AI tokens and advised to limit their use.
- In mid-June, the Army CIO pool of unlimited AI tokens was exhausted, leading to re-establishment of limits.
- The Army's generative AI platform, Ask Sage, has an annual subscription of 100 million tokens which are quickly consumed by users.
- Employees were encouraged to maximize their use of approximately 200,000 tokens allocated per month for AI tasks.
The Army uses Ask Sage, a generative AI platform that works with several large language models to assist in various tasks. The use of AI has been promoted within the Army, yet the rapid consumption of tokens highlights challenges in managing and utilizing these resources effectively.
Read the original ↗