From The New Stack · 40 stories
OpenAI Expands Daybreak Initiative with $1 Billion for Critical Infrastructure Cyber Defense
OpenAI announced Daybreak for Frontline Defenders, an expansion of its existing Daybreak initiative, committing $1 billion to help critical infrastructure sectors like power, water, and banking use frontier cyber AI for defense. This initiative provides subsidized access to AI models, training, and support to cyber defenders protecting essential services globally and within the United States.
Z.ai Releases GLM-5.3 with Enhanced Coding and Cybersecurity Capabilities
Chinese AI startup Z.ai launched GLM-5.3, an updated language model with significant improvements in long-horizon coding and cybersecurity capabilities, which reportedly identified a serious vulnerability in Cursor. The model's advancements come from scaling post-training rather than a new base model, highlighting the potential for existing models to gain new capabilities through further training.
Kubernetes Extends Reach to Desktop Infrastructure, Highlighting Database Management Challenges
Kubernetes is being considered for managing desktop infrastructure, traditionally separate from cloud-native models. This shift aims to unify operational practices and lower costs related to outdated virtual desktop systems. However, while Kubernetes simplifies deployment, it also exposes complexities in managing databases, requiring expertise beyond standard DevOps skills.
Stripe Reportedly Acquires AI Gateway Startup OpenRouter for Over $7 Billion
Stripe has reportedly finalized a deal to acquire OpenRouter for more than $7 billion, according to Bloomberg. OpenRouter provides a single access point for customers to select and use various AI models, preventing vendor lock-in.
Grok AI Vulnerable to Data Exfiltration via Encrypted Malicious Instructions
Researchers discovered a new prompt injection attack against Grok that uses encrypted malicious instructions to bypass guardrails and exfiltrate user data. This method exploits the LLM's inability to distinguish between trusted user input and harmful content, allowing it to steal chat data and personal information.
DeepSeek Updates V4-Flash Model, Raises API Prices, and Launches V4-Pro
DeepSeek has released DeepSeek-V4-Flash-0731, an updated version of its V4-Flash model, now in public beta, which maintains its architecture but shows performance gains through post-training. Concurrently, DeepSeek launched its V4-Pro model with agent upgrades and introduced new API pricing with peak and off-peak rates, effective August 16, 2026, which will increase costs for both V4-Flash and V4-Pro models.
Apple Implements Bug Report Caps Due to Surge in AI-Generated Submissions
Apple has introduced a cap on open security reports and a 30-day cool-off period for submissions to its bug bounty program, effective June. This change was made in response to a significant increase in AI-assisted reports, many of which were not genuine vulnerabilities, leading to review teams being overwhelmed. The new policy impacted Italian cybersecurity company Bynario, which used GPT-5.5 to find a critical macOS bug (CVE-2026-43760) but was initially unable to report it due to reaching the submission limit.
Google DeepMind's Gemini Robotics 2 Enables Whole-Body Control for Humanoid Robots
Google DeepMind has released Gemini Robotics 2, an updated AI model that allows humanoid robots to control their entire bodies, from feet to fingertips. This advancement enables robots to perform a wider range of actions and more complex dexterous tasks, moving towards physical Artificial General Intelligence (AGI). The update includes new sub-models, with one, Gemini Robotics ER 2, now available for developers.
Google DeepMind CEO Demis Hassabis Proposes U.S.-Led Global AI Regulation Body
Demis Hassabis, CEO of Google DeepMind, has suggested the formation of a U.S.-led global AI watchdog to regulate advanced AI models. This proposed body would assess the safety of AI systems before their release and manage risks associated with emerging technologies like artificial general intelligence. Hassabis emphasizes the need for urgent regulation as AI developments pose increasing cybersecurity and biosecurity threats.
Google Develops 'Frozen v2' AI Chip to Enhance Gemini Model Efficiency by 2028
Google is developing a new AI chip, called 'Frozen v2', designed to improve the efficiency of its Gemini models. Projected to be 6-10 times more efficient than current chips, it aims to address AI compute shortages and reduce dependency on external hardware. This development reflects a shift towards in-house chip production for AI workloads.
Linus Torvalds Backs AI Tools in Linux Development Amid Community Debate
Linus Torvalds has confirmed his support for using AI tools in Linux development, amid controversy over AI-generated code. He emphasized Linux will not take an anti-AI stance and suggested dissenters fork the project or leave. This stance reflects AI's growing role in open-source projects.
Multiverse Computing Releases Quasar 438B, a New Enterprise AI Reasoning Model
Multiverse Computing has released Quasar 438B, its first large reasoning model, designed for enterprise agents and coding. The model scores 43 on the Artificial Analysis Intelligence Index, making it the highest-scoring European model, and demonstrates competitive speed for its class.
PhiloLabs AI Agents Reconstruct Virtual Union Square for $33, Identify Visual Errors
PhiloLabs used Claude Fable 5.1 AI agents to reconstruct a 3D virtual model of San Francisco's Union Square for $33 in API calls. The experiment highlighted the agents' ability to not only build functional code but also identify visual inaccuracies using Playwright for comparison against real-world images.
Debian Considers Banning AI-Assisted Code Contributions to Maintain Stability
The Debian project is debating proposals regarding the use of LLM-assisted contributions, including an outright ban, to preserve its reputation for stability. This decision could significantly impact how developers contribute to one of the most widely used open-source operating systems.
X issues cease-and-desist letters to Nitter, an open-source X content viewer
X has sent cease-and-desist letters to the open-source project Nitter, demanding the shutdown of its instances and repository due to alleged unlawful scraping and API circumvention. This legal action follows previous technical attempts by X to disable Nitter, which allowed users to view X posts without ads or tracking. The Nitter.net instance is now offline, and development has ceased as the creator seeks legal advice, impacting users who relied on the service for ad-free X content viewing.
Slack launches collaborative coding channels with AI agent integration
Slack has introduced Slack Code, a new feature that provides dedicated channels for teams to collaborate on coding tasks with AI agents. This allows developers to work with AI agents like Anthropic's Claude or Cognition's Devin directly within Slack, streamlining the development workflow and providing visibility into the coding process.
Apple Partners with Alibaba for China-Specific AI, Diverging from Global Strategy
Apple will use Alibaba's Qwen models to power new Siri AI features in China, a departure from its global partnership with Google's Gemini models. This decision is driven by Chinese government regulations requiring local AI partnerships, which may result in different privacy protections for users in China compared to other regions.
White House Accuses Chinese AI Firm Moonshot of Siphoning Anthropic's Fable 5 Data
White House official Michael Kratsios accused Chinese AI startup Moonshot of deceptively extracting data from Anthropic's Fable 5 model to train its Kimi K3 system. This accusation signals a potential shift in US policy, framing large-scale model distillation as theft of American technology, which could lead to tighter API restrictions and expanded export controls on AI chips. Treasury Secretary Scott Bessent warned of sanctions against Chinese AI companies for intellectual property theft.
Block Launches Buzz: Open-Source Team Chat Integrating AI Agents and Git Hosting
Jack Dorsey and Block have launched Buzz, an open-source collaboration platform that merges team communication, AI agents, and Git hosting, built on the Nostr messaging protocol. Buzz provides a Slack-like interface where AI agents and humans interact, offering customizable workflows and cryptographic identities for all participants. This release potentially reduces reliance on multiple existing tools like Slack and GitHub.
Apache Spark 4.2 Launches with Native Vector Search and Governed Metrics
Apache Spark 4.2 introduces significant new features including native vector search and governed metrics. This release enhances Spark's capabilities for AI workloads and reduces reliance on separate vector databases, potentially transforming enterprise data processing workflows.
Thomson Reuters Develops Proprietary AI Model for Legal and Tax Work
Thomson Reuters has developed its own AI model, named Thomson, for legal, tax, and compliance tasks, trained on its proprietary content. This model will power specific features within products like CoCounsel, while the company continues to use third-party models like Anthropic's Claude for other functionalities. This approach allows companies with extensive proprietary data to create specialized AI without building a foundation model from scratch.
Mistral AI Expands Infrastructure, Commits to 1 Gigawatt European Compute by 2030
Mistral AI announced a three-part expansion of its infrastructure business, including regional inference endpoints, a "Priority Tier" with a 99.5% uptime guarantee, and commitments to build 1 gigawatt of European compute by 2030. Five European companies have made multi-year compute commitments to underwrite 200 megawatts by 2027. The company will also host third-party open models, starting with GLM-5.2 from Z.ai, to provide a unified platform for enterprises.
Manus to become independent after Chinese regulators unwind Meta's $2 billion acquisition
Manus, an AI startup, will resume independent operations after Chinese regulators mandated Meta unwind its $2 billion acquisition. This decision, issued by the National Development and Reform Commission in April, requires Manus to separate from Meta and has led to a data backup requirement for some users.
Amazon EKS Introduces Kubernetes Version Rollback and Extended Support
Amazon EKS now supports Kubernetes version rollbacks, allowing users to revert a cluster's control plane to its previous Kubernetes version within seven days of an upgrade. This feature provides a safety net for quick recovery from problematic updates. Additionally, EKS has introduced Extended Support for Kubernetes versions, making each version available for 26 months.
Nscale Acquires Anyscale for $1.65 Billion to Expand AI Compute Stack
British AI neocloud Nscale has acquired software startup Anyscale for an estimated $1.65 billion. This acquisition integrates Anyscale's AI workload scaling capabilities, built around the open-source Project Ray framework, into Nscale's infrastructure, allowing Nscale to offer a more comprehensive AI compute stack from energy and data centers to workload management.
IBM and Partners Address Quantum Computer Verification Challenges with New Methods
IBM and its partners have published three preprint papers detailing new methods to verify results from quantum computers, even as these machines perform calculations that exceed the capabilities of classical computers. This development addresses the challenge of trusting quantum computation results when classical verification is no longer possible, which is crucial for demonstrating quantum advantage.
Cursor Launches India-Specific Subscription Plan Ahead of SpaceX Acquisition
AI coding startup Cursor introduced "Cursor Start," a new subscription plan priced at ’649 (approximately $7) per month, specifically for the Indian market. This localized pricing aims to expand Cursor's user base in India, which is already its third-largest market and home to a high concentration of power users, ahead of its anticipated acquisition by SpaceX.
Model Context Protocol Receives Major Update, Adopting Stateless Architecture
The Model Context Protocol (MCP), an open standard for AI agents, has received its largest update since its launch, transitioning to a fully stateless architecture. This revision removes sessions and the initialization handshake, deprecates three core features, and graduates two capabilities into official protocol extensions, simplifying deployment for large-scale enterprise AI agent applications.
Cloudflare open-sources 'pvcli' debugger for OHTTP and MASQUE privacy protocols
Cloudflare has open-sourced "privacy-client" (pvcli), a command-line debugger for Oblivious HTTP (OHTTP) and MASQUE protocols, under an Apache 2.0 license. This tool addresses the difficulty of troubleshooting privacy services like Apple's iCloud Private Relay and Microsoft's Edge Secure Network VPN, which split trust across multiple operators to prevent any single entity from linking user identity to online activity.
DoorDash Introduces Command-Line Interface to Streamline Ordering
DoorDash launched a limited beta of dd-cli, a command-line interface for macOS developers in the US and Canada to enable AI-mediated food ordering, enhancing integration with autonomous AI agents.
Runway Introduces Solaris, an AI Model for Real-Time Generative User Interfaces
Runway launched Solaris, its first Interface World Model, which generates user interfaces frame by frame in real time based on user interactions. This model aims to eliminate the need for translating designs into code by making the visual interface the application itself, allowing for direct interaction with generated scenes.
Meta launches Muse Code out of beta with new features and subscription tiers
Meta has officially launched its coding agent, Muse Code, out of beta, introducing three new subscription plans ranging from $5 to $50 per month. The launch includes new capabilities like inter-session messaging, Workflows for orchestrating agents, and a rewind feature. This move provides predictable monthly pricing for the coding agent, which is powered by Meta's Muse Spark 1.2 model.
Perplexity's Personal Computer AI Agent Now Available on Windows
Perplexity has expanded its Personal Computer feature, an agentic AI designed for multi-step tasks, to Windows 10 and 11. This feature, previously Mac-exclusive, allows the AI to access local files, native applications, and cloud services like OneDrive and Google Drive, requiring a paid Perplexity subscription.
Google Releases TimesFM-3, a New Time-Series Forecasting Model, Under Non-Commercial License
Google launched TimesFM-3, a 330-million-parameter time-series forecasting model trained on over a trillion data points, now available on Hugging Face with a non-commercial license. This model is Google's first natively pre-trained for multivariate time series and zero-shot generalization, outperforming previous models in benchmarks. Its non-commercial license restricts immediate business use, but it advances the field of time-series forecasting.
Git Worktrees Lack Isolation for AI Agents, Creating Security and Infrastructure Challenges
Git worktrees, often used for parallel execution of AI coding agents, do not provide true isolation, allowing agents to manipulate repository state or install malicious hooks outside their designated worktree. This lack of isolation poses security risks and creates bottlenecks in existing runtime infrastructure, as multiple agents generate changes in parallel, overwhelming shared resources like CI queues and staging environments.
Amazon ECS Express Mode simplifies container deployment on AWS Fargate
Amazon Web Services (AWS) has launched ECS Express Mode, a new interface for Amazon Elastic Container Service (ECS) that simplifies the deployment of containerized applications. This mode automates the setup of load balancers, networking, IAM roles, and scaling policies, allowing developers to deploy an HTTPS service on Fargate with just a container image and two IAM roles.
LM Studio develops Auto Review and Shell Judge to secure AI coding agent commands
LM Studio introduced Auto Review and Shell Judge to enhance the security of commands executed by its Bionic AI coding agent. These tools analyze shell commands to prevent potentially dangerous operations, particularly when variables are involved, by parsing command structure and understanding tool-specific interpretations.
Google DeepMind introduces double-blind evaluation for AI models using confidential computing
Google DeepMind has developed a double-blind evaluation method for AI models, allowing testing without either the model provider or the evaluator seeing the other's proprietary data. This method addresses benchmark leakage concerns by using confidential computing to protect both model weights and test questions during evaluation. The approach aims to provide a more secure and unbiased way to assess AI model performance.
Simular's Sai Agent Achieves 73% Success on OSWorld 2.0 Benchmark
Simular's Sai agent scored 73% on the OSWorld 2.0 benchmark, outperforming OpenAI's GPT-5.6 Sol and Anthropic's Opus 5. This result indicates progress in AI agents handling complex, real-world professional tasks at a lower operational cost.
Perplexity introduces Portable Computer, separating AI reasoning from deterministic execution
Perplexity released Portable Computer, a local-first version of its Computer agent, which separates probabilistic AI reasoning from deterministic software execution. This architectural change allows for improved reliability and performance in AI agents by using code for control and policy enforcement, rather than relying solely on larger AI models for orchestration.