OpenAIModel
OpenAI Updates GPT-Rosalind
GPT-Rosalind now features improved tools for biological reasoning and genomics analysis. The model also offers expanded expertise in medicinal chemistry and experimental workflows.
Read the source →Thursday, June 4, 2026
30 stories
AnthropicResearch
Anthropic Analyzes Year of AI-Driven Cyber Attacks
Anthropic has released its findings after tracking AI-powered cyber threats for one year. The study examines the patterns and evolution of these digital risks.
Read the source →
Google AIProduct
Google Enhances Second-Hand Shopping via AI
Google is introducing AI-powered features within Search and Shopping to help users find vintage items. These tools are designed to simplify the process of discovering second-hand goods.
Read the source →
Hugging FaceResearch
Expanding Direct Preference Optimization Applications
Researchers are exploring the use of Direct Preference Optimization for tasks outside of conversational AI. This work examines how the technique can be applied to a broader range of model objectives.
Read the source →
NVIDIAResearch
NVIDIA Advances Robotic Grasping and Autonomous Driving
NVIDIA Research is developing systems to improve how robots handle unfamiliar objects. The work also focuses on enhancing the safety and scalability of autonomous vehicle training.
Read the source →
AWS Machine LearningProduct
AWS Introduces Bedrock Ops Alert for AI Monitoring
Amazon Bedrock Ops Alert is a new three-tier automated system designed to manage operational issues and notify SRE teams. It features dynamic alarm adjustment and automatic case creation to reduce duplicate support requests.
Read the source →
TechCrunchFunding
Lovable Expands Partnership with Google Cloud
Lovable has entered into a multiyear agreement to significantly increase its infrastructure on Google Cloud. The deal also provides the company with broader access to Anthropic's Claude models.
Read the source →
The VergeProduct
Google's Spark Agent Raises Privacy Concerns
Reviewers found Google's new Gemini-powered agent, Spark, to be alarmingly efficient. The tool demonstrated an unsettling ability to access personal details about the users' family and pets without prior disclosure.
Read the source →
WiredPolicy
AI Leaders Push for Stricter DNA Monitoring
OpenAI and Anthropic have joined other industry experts in urging government officials to enhance the oversight of synthetic DNA. The goal is to prevent the use of artificial intelligence in creating biological weapons.
Read the source →
The DecoderModel
Google DeepMind Launches Gemma 4 12B
This open-source multimodal model can handle text, audio, and images on laptops equipped with 16 GB of RAM. Despite its smaller size, it performs nearly as well as the 26B version and is available for commercial use via an Apache 2.0 license.
Read the source →
Ars TechnicaPolicy
Google Faces New UK Requirements for AI Search Results
Regulators have directed Google to provide more transparent linking within its AI Overviews. Additionally, publishers in the UK must be given the option to prevent their content from being used.
Read the source →
OpenAIProduct
Wasmer Accelerates Node.js Runtime Development via Codex
Wasmer utilized Codex and GPT-5.5 to create a Node.js runtime designed for edge computing. This approach reduced the development timeline from months to weeks, increasing production speed by 10 to 20 times.
Read the source →
AnthropicProduct
Anthropic Expands Its Claude Partner Network
Anthropic has launched a new Partner Hub and a dedicated Services Track. These additions are designed to enhance the existing Claude Partner Network.
Read the source →
Hugging FaceProduct
MCP Tools Integrated into Reachy Mini
Reachy Mini has been updated to include support for Model Context Protocol (MCP) tools. This integration expands the capabilities of the robotic system.
Read the source →
NVIDIAResearch
NVIDIA Introduces New Tools for Physical AI
NVIDIA presented new agent skills at CVPR designed to accelerate the creation of robotics, vision AI, and self-driving systems. These tools focus on improving the entire development workflow, including scene reconstruction and the generation of rare edge-case scenarios.
Read the source →
AWS Machine LearningProduct
NEXUS Tabular Model Joins SageMaker JumpStart
Fundamental's NEXUS model for tabular data is now accessible via Amazon SageMaker JumpStart. Users can deploy the tool to generate predictions from corporate datasets.
Read the source →
OpenAIProduct
Endava Integrates AI Agents into Software Development
Endava is leveraging Codex and ChatGPT Enterprise to speed up the delivery of software. The company aims to automate internal workflows and establish an AI-centric corporate culture.
Read the source →Hugging FaceResearch
Customizing Nemotron 3.5 ASR
Hugging Face has shared a guide on fine-tuning the Nemotron 3.5 automatic speech recognition model. The process allows users to adapt the system to specific accents, languages, or professional domains.
Read the source →AWS Machine LearningModel
NVIDIA Nemotron 3 Ultra Hits Amazon SageMaker
Amazon SageMaker JumpStart now supports the deployment of NVIDIA Nemotron 3 Ultra. This reasoning model is designed to reduce costs by 30% and significantly increase inference speeds for agentic AI tasks.
Read the source →
TechCrunchProduct
Meta Launches AI Helper for Facebook Creators
Facebook now features an AI assistant designed to help content creators analyze their performance. Instead of manually studying data charts, users can ask the tool for posting schedules and audience sentiment.
Read the source →
The VergeOther
TSMC Faces Capacity Constraints Amid AI Surge
The world's largest chip manufacturer is having difficulty fulfilling orders from U.S. clients. Despite expanding its American facilities, CEO C.C. Wei noted that current demand is exceeding the company's production capabilities.
Read the source →
WiredFunding
Bezos Invests in Biological AI Research
The startup Flourish has received $500 million in backing and is valued at $2.5 billion. The company aims to transform artificial intelligence by studying actual neurons to uncover the brain's primary operating logic.
Read the source →
MIT Technology ReviewPolicy
Judges Face Surge in AI-Written Legal Filings
Legal professionals are dealing with an increase in court documents produced by artificial intelligence. This trend is particularly evident among individuals representing themselves who cannot afford legal counsel.
Read the source →
The DecoderProduct
ChatGPT Enhances User Memory System
OpenAI has updated ChatGPT to create organized narrative profiles based on user interests and work rather than simple lists. This improvement has increased the accuracy of maintained information from 52.2% to 75.1%.
Read the source →
OpenAIProduct
ChatGPT Enhances Long-Term Memory
OpenAI has implemented a new system that allows ChatGPT to recall user preferences. This update ensures that information remains consistent and relevant throughout different chats.
Read the source →
Hugging FaceResearch
Introduction of EVA-Bench Data 2.0
A new evaluation framework called EVA-Bench Data 2.0 has been released. It features 213 different scenarios across three domains and incorporates 121 tools.
Read the source →
TechCrunchProduct
Hello Robot Launches New Home Assistant
The California-based company has introduced the fourth version of its Stretch robot. This device is designed to provide assistance within residential settings.
Read the source →
The VergePolicy
The Demand for Better AI Content Filtering
Major social media platforms have started implementing labels to identify synthetic media. However, there is a growing push for more robust tools to help users filter out AI-generated material entirely.
Read the source →
The DecoderResearch
Human Intervention Hinders AI Cost Savings
A Bain survey of 951 firms reveals that nearly 40% failed to reach their AI-driven financial goals. This gap is partly attributed to a lack of fully autonomous agents, which only 7% of these companies have actually implemented.
Read the source →
Ars TechnicaPolicy
AI Testing Plans Face Resource Shortages
Donald Trump's initiative to evaluate AI models is facing criticism for being superficial. Opponents argue the plan is hindered because DOGE has significantly reduced the staffing of US security teams.
Read the source →