Ai · Top stories
Gradio Introduces gr.Workflow for Building AI Pipelines with Visual Interfaces and REST APIs
Gradio has launched gr.Workflow, a new feature that allows users to define AI pipelines as a graph of typed nodes, providing a drag-and-drop canvas for interaction. This development simplifies the creation and deployment of multi-step AI applications, making them accessible as both visual interfaces and REST APIs, and enabling one-command deployment to Hugging Face Spaces.
Instinct AI Assistant Raises Privacy and Security Concerns During Private Testing
Instinct, an AI personal assistant from Spear Street Technology, is drawing privacy and security concerns from testers due to its broad data access and terms of service. The AI connects to user applications and devices, and its terms grant it extensive rights to user data, including for model training. These concerns highlight the trade-offs involved in granting AI agents significant access and autonomy.
Robotaxi Expansion Faces Increased Regulatory and Public Pushback Across US Cities
The expansion of robotaxi services is encountering growing resistance from local governments, labor unions, and safety advocates, leading to stalled legalization efforts and calls for stricter regulations in various US cities. This pushback highlights the challenges policymakers face in overseeing autonomous vehicle technology as it moves from development to widespread public use.
Goldman Sachs Partner Warns AI Could Degrade Bankers' Reasoning Skills
A Goldman Sachs partner leading an AI project, Chris Churchman, warned that relying too heavily on AI could diminish the critical thinking abilities of future financiers. He stated that outsourcing reasoning to AI models risks "cognitive atrophy" among bankers, similar to how modern inventions have affected navigation and memorization skills. This concern highlights a potential long-term impact on talent development within the financial industry as AI automates more routine tasks.
Apple begins testing AI shopping assistant in Apple Store app for select users
Apple is rolling out an AI-powered shopping assistant in its Apple Store app to select users as part of an "Early Preview." This integration allows users to interact with a virtual assistant for product inquiries, order status, and trade-in values, marking Apple's expansion of AI features into its retail experience.
Microsoft Introduces AI Governance Architecture for Runtime Enforcement and Continuous Evaluation
Microsoft has released an AI governance architecture that shifts focus from policy documents to runtime enforcement, continuous evaluation, and observability for AI applications in production. This framework aims to ensure governance requirements are enforced and auditable during AI system operation, addressing the need for safer AI systems as organizations deploy AI workloads.
Alibaba shares drop 10% after $10.2 billion share placement to fund AI development
Alibaba's shares fell by 10% in Hong Kong following the announcement of an $10.2 billion share placement aimed at funding its AI capabilities and infrastructure. This move comes after the company reported a 75% drop in quarterly profit due to significant AI spending.
Inherent's Faraday AI Agent Outperforms Larger Models in Scientific Paper Replication
Inherent, an AI lab founded by DeepMind alumni, announced that its new AI agent, Faraday, successfully replicated scientific paper findings, outperforming larger models from Anthropic and OpenAI. This achievement demonstrates the potential for smaller, more efficient AI models in scientific research tasks, highlighting a different approach to AI development through reinforcement learning.
Hollywood Creatives Train AI Models to Replicate Their Skills Amid Industry Downturn
Hollywood writers, directors, and producers are taking gig work to train AI models to perform tasks like screenplay writing and scheduling, earning between $12 and $200 per hour. This trend is occurring during a significant downturn in entertainment industry jobs, with some creatives expressing concern that they are contributing to the automation of their own professions.
GPT-5.6 Sol input token pricing reduced by 20%, output by 33%
The GPT-5.6 Sol model now costs $4 per million input tokens and $20 per million output tokens, representing a 20% reduction in input pricing and a 33% reduction in output pricing. This pricing change makes the frontier model in the GPT-5.6 family more accessible for developers and applications, impacting the cost-effectiveness of integrating advanced AI capabilities.
Waymo Submits Redacted Documents to NHTSA for Child Collision Investigation
Waymo has provided redacted documents to the National Highway Traffic Safety Administration (NHTSA) as part of an investigation into a January incident where a robotaxi struck a child. This submission indicates progress in the regulatory inquiry concerning autonomous vehicle safety protocols, particularly around schools.
Florida Sues OpenAI and Sam Altman, Seeks Public Nuisance Ruling for ChatGPT
Florida has filed a lawsuit against OpenAI and its CEO Sam Altman, seeking a court ruling that classifies ChatGPT as a public nuisance and alleging violations of state consumer protection laws. OpenAI removed the case to federal court, arguing federal jurisdiction due to a Children's Online Privacy Protection Act (COPPA) claim, while Florida is fighting to keep the case in state court.
Law School Prohibits AI Use for Academic Work to Preserve Core Legal Skills
A law school has implemented a policy, effective Summer 2026, prohibiting students from using AI for conceptualizing, outlining, drafting, revising, translating, or editing any work submitted for credit. This policy aims to ensure students develop essential cognitive skills for legal practice and uphold ethical obligations, while allowing instructors to deviate for courses specifically designed to teach AI fluency.
Runlayer and Rippling File Lawsuits Against Each Other Over AI Product
AI startup Runlayer has sued HR software company Rippling, alleging trade secret misappropriation and breach of contract related to Runlayer's Model Context Protocol (MCP) gateway product. Rippling subsequently filed a counter-lawsuit, accusing Runlayer of infringing on three of its patents. The dispute stems from a product trial where Rippling evaluated Runlayer's technology before developing its own competing MCP server.
Ramp Launches AI Model Router Service for LLM Access and Management
Ramp, a corporate expense management platform, has launched "Router," an AI model routing service that allows users to access and switch between various large language models via an API. This service aims to tap into the AI inference market and integrate with Ramp's existing AI token usage monitoring and spend management products.
AlloyDB ScaNN now scales vector search to 10 billion vectors with new architecture
AlloyDB's ScaNN index has been updated with a new four-level tree architecture, enabling it to scale vector search to 10 billion vectors. This enhancement addresses memory and computational challenges in large-scale AI applications, improving efficiency for enterprise workloads.
Serval launches Catalyst AI agent for automated IT issue identification and resolution
Serval has made its Catalyst AI agent generally available, allowing it to automatically identify recurring IT issues and build automation workflows to resolve them. Catalyst also creates background agents that proactively inspect systems for problems and propose fixes before tickets are filed, differentiating it from other AI-assisted workflow tools.
Adobe Firefly AI audio generation tools, including music and voiceover, are now generally available
Adobe has made its Firefly AI audio generation tools, including Generate Music, Generate Speech, and Generate Sound Effects, generally available. These tools allow users to create fully licensed music tracks, natural voiceovers, and custom sound effects directly within Firefly. This release addresses common challenges for video creators, musicians, and marketers related to music licensing costs and copyright risks, integrating audio creation into existing workflows.
Google Research Develops PhotoScan for Insulin Resistance Prediction from Smartphone Imagery
Google Research has developed PhotoScan, a deep learning method that estimates body composition from smartphone photos to predict insulin resistance with accuracy comparable to DXA scans in a clinical research setting. This development offers a more accessible and less invasive alternative to traditional methods for early detection of metabolic disease drivers.
Binance Launches Agent OS Platform for AI-Driven Crypto Trading
Binance introduced Agent OS, a new platform enabling AI agents to analyze markets and execute cryptocurrency trades on behalf of users. This development integrates autonomous AI directly into financial management on the world's largest crypto exchange, giving users control over agent permissions and trading limits.
New adaptation method ports task knowledge across frozen models with high accuracy
A new method allows task adaptations to be learned once in a base-agnostic form and then ported to new frozen models by refitting a thin per-base alignment. This approach recovers approximately 98% of per-task LoRA's performance on unseen models within the same family and 94% across different families.
Google Search integrates new AI-powered learning tools for interactive visuals and practice quizzes
Google Search is rolling out new AI-powered features to assist with learning, including interactive visuals for complex concepts and customizable practice quizzes. These additions aim to provide students with more dynamic and personalized educational resources directly within Search.
OpenAI's Codex integrates asynchronous messaging for developer interaction
OpenAI has merged `send_user_message_async` into the public Codex repository, allowing the AI coding agent to continue working on independent tasks while waiting for developer input. This change improves efficiency by preventing Codex from idling, enabling it to ask questions and proceed with other work concurrently.
Claude Opus 4.8 and 5.0 criticized for incoherent and toxic language style
Users report that Claude Opus 4.8 and 5.0 models exhibit a verbose, jargon-heavy, and over-stylized writing style, making responses difficult to understand. This issue, widely discussed on Reddit, affects the chat/web UI, Claude Code, and Fable 5, leading some users to consider alternative AI providers.
China Relocates AI Data Centers to Rural Provinces for Energy and Land
China is moving AI data center complexes to rural provinces like Guizhou, following a "Eastern Data, Western Computing" strategy. This move aims to utilize surplus energy and land, with companies like Huawei and Tencent building infrastructure in these areas, contrasting with data center development challenges in the US.
MicroGPT in pure C achieves 10M tokens/sec inference on Apple M5
A character-level transformer, MicroGPT, implemented in pure C with no external dependencies beyond libc, achieved over 10 million tokens per second inference speed on an Apple M5 chip. This implementation demonstrates efficient GPT training and inference in a minimal C environment, highlighting performance capabilities on modern hardware.
Vivodyne Develops Robotic Labs to Generate Human Tissue Data for AI Drug Discovery
Biotech startup Vivodyne introduced HIVE, modular robotic labs capable of growing and monitoring 20 types of human tissue to generate causal biological data. This initiative addresses a data gap in AI drug discovery, which currently relies on less relevant animal or single-cell data, hindering the development of effective human treatments.
Taiwan to Distribute $314 Cash Payout to Residents Funded by AI Server Boom
Taiwan's President Lai Ching-te announced a plan to distribute NT$10,000 (approximately $314 USD) to every resident in 2027, funded by a surge in government revenue attributed to global demand for AI servers. This initiative aims to share the economic benefits of the AI industry's growth across the population, as Taiwan's economy experiences significant expansion driven by AI-related exports.
Block open-sources Berd, a desktop AI agent workspace for multiple models
Block has open-sourced Berd, a desktop application designed to provide a unified environment for working with various AI agents and models. This release allows developers and companies to use, modify, and redistribute the tool, addressing the fragmentation issues Block employees faced when using different AI interfaces.
Robin Williams' Children Revive Instagram to Combat AI Misuse of Likeness
Robin Williams' children have reactivated their late father's Instagram account to create a trusted space for his authentic legacy and to counter the unauthorized use of his voice and likeness by AI. This action highlights ongoing concerns within the entertainment industry regarding AI-generated content and the challenges of protecting public figures' digital identities.
Nova3D Generates Programmable 3D Assets as Executable Blender Source Code
Researchers developed Nova3D, a system that generates 3D assets as executable Blender source code, treating the compiled mesh as an artifact rather than the asset itself. This approach allows for semantic handles at generation time, enabling downstream systems to inspect, measure, edit, and animate the generated 3D objects.
AI-generated bills create errors and workload for US House legislative lawyers
The US House Office of Legislative Counsel is experiencing increased workload due to errors in AI-generated legislative proposals submitted by representatives' offices. These proposals, drafted using public AI tools like ChatGPT, contain incorrect terms, citations, and miss crucial nuances, requiring extensive review and rewriting by lawyers. This development highlights the challenges of integrating general AI tools into complex legislative processes and raises concerns about congressional staffers' understanding of the bills they introduce.
Google Acquires Spirit Aviation Holdings Data for AI Development
Google LLC won a bankruptcy auction for $10 million to acquire deidentified business data, software code, and operational records from the defunct Spirit Aviation Holdings Inc. This acquisition provides Google with a large dataset, including emails and chat records, which it plans to use to improve its artificial intelligence capabilities, particularly for its travel booking services.
Federal Court Upholds Judicial Immunity for Judge Allegedly Using AI for Rulings
A U.S. federal court ruled that judicial immunity protects a state judge from a civil lawsuit, even if the judge allegedly used AI to make a judicial decision. The ruling did not determine whether the judge actually used AI or if such use is legally permissible, but rather affirmed the broad scope of judicial immunity. This decision clarifies the limits of challenging judicial actions, even in novel circumstances involving AI.
OpenAI, AWS, GitHub, Microsoft, and Cursor back Agent Plugins 1.0.0 standard
OpenAI, AWS, Cursor, GitHub, and Microsoft have endorsed Agent Plugins 1.0.0, a new open standard for AI agent plugins initiated by Vercel. This standard provides a common, portable package format for reusable components, allowing AI agents to connect to external tools and services consistently across different clients and environments.
Higgsfield Raises $400M Series B, Valuing AI Video Platform at $5.4B
Higgsfield, an AI image and video creation platform, secured $400 million in Series B funding, increasing its valuation to $5.4 billion just eight months after its previous round. This capital will support hiring, product development, and the significant compute resources required for AI video processing, as the company expands its enterprise adoption.
Roboflow Launches Playground for Side-by-Side Comparison of 30+ Computer Vision Models
Roboflow introduced Playground, a new tool that allows users to compare the outputs of over 30 zero-shot computer vision models side-by-side using the same image and prompt. This platform removes the need for individual API provisioning or infrastructure setup, simplifying the evaluation process for various vision tasks like object detection, classification, and OCR.
ByteDance and MPA Sign Agreement to Protect Hollywood IP in AI Models
ByteDance and The Motion Picture Association (MPA) signed a memorandum of understanding to establish a framework for protecting Hollywood intellectual property from infringement by ByteDance's AI models. This agreement follows a cease-and-desist letter from the MPA in February and ByteDance's subsequent commitment to strengthen IP safeguards, impacting AI models used across TikTok, CapCut, and Dreamina.
Alibaba Sells Gaming Unit Lingxi Games for Over $1.5 Billion to Fund AI Development
Alibaba is selling its game development unit, Lingxi Games, to private equity firm Trustar Capital for at least $1.5 billion. This divestment allows Alibaba to reallocate capital towards its expanding AI infrastructure program, which has exceeded its initial budget.
Expert Witness Used ChatGPT to Draft Report Defending 3M in Deadly Explosion Lawsuit
An expert witness hired by 3M in a lawsuit concerning a fatal 2020 explosion used ChatGPT to write significant portions of his expert report, including prompts asking the AI to defend 3M and state it was "0% at fault." This incident highlights the increasing presence of AI in legal proceedings, extending beyond AI-generated briefings to expert testimonies, and raises concerns about the discoverability and potential embarrassment of AI prompts in court cases.