Apple plans to launch an AI-overhauled Siri next year, using a test app called Veritas to evaluate its enhanced capabilities, including personal data search and in-app actions like photo editing. The updated Siri combines Apple’s internal LLM with an external model. Delays due to engineering challenges pushed the release from spring 2025.

Microsoft releases .NET Aspire 9.5, supporting .NET 8, 9, and 10 RC1. Key features: *aspire update* preview for auto-scanning and applying updates, single-file AppHost, enhanced *aspire exec*, Dev Containers SSH support, AI-driven dashboard telemetry, improved integrations (OpenAI, GitHub Models, Dev Tunnels), YARP static files, Azure Kusto preview, Redis/RabbitMQ auto-activation, and new deployment capabilities. Breaking changes include InteractionInput API name requirement and notification sys

Oracle introduces MySQL AI in Enterprise Edition, offering vector storage, LLM integration, and natural language querying. Concerns arise over open-source version's future and potential vendor lock-in.

Meta's Code World Model (CWM-32B), a 32B-parameter AI, revolutionizes code generation and debugging by simulating code logic and execution behavior, offering semantic understanding and causal reasoning. Trained through pre-training, mid-training, and reinforcement learning, CWM observes code execution, predicts outcomes, and integrates with virtual agents. With applications in debugging, intelligent agents, and domain-specific tasks, CWM acts as a "neural debugger," enhancing developer productiv

Google DeepMind released Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, enabling robots to reason, plan, and generalize across tasks like sorting laundry, packing based on weather, and recycling. The models combine online search, perception, and multi-step planning, marking progress toward adaptable, general-purpose robotics.

Google DeepMind upgrades AI models Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, enabling robots to handle complex, multi-step tasks, perform online searches, and transfer learned skills across different robots.

Spotify updates AI policies to label AI-generated music, block unauthorized voice cloning, and introduce a spam filter to combat misuse, while supporting responsible AI use in music creation.

Databricks partners with OpenAI in a $100M multi-year deal to integrate OpenAI's models, including GPT-5, into its data platform and new AI product Agent Bricks, aiming to attract enterprise clients. Agent Bricks enables businesses to build custom AI apps using various models, with GPT-5 as a flagship offering. The agreement guarantees OpenAI revenue regardless of performance, marking Databricks’ strategic push into enterprise generative AI.

OpenAI introduced GDPval, a benchmark evaluating AI models' performance against human professionals across 44 jobs in nine key U.S. industries. Early tests show GPT-5 and Claude Opus 4.1 nearing expert human quality, with win rates of 40.6% and 49%, respectively. While not yet replacing human labor, the tool tracks AI progress toward AGI and real-world task proficiency.

Grok, xAI's AI chatbot, is now available to the U.S. federal government for 42 cents per year under an 18-month agreement, offering technical support and undercutting competitors like OpenAI and Anthropic. The deal, part of broader Pentagon contracts, follows internal government pushes and reflects Musk's influence through appointees in key agencies.

OpenAI launches ChatGPT Pulse, delivering personalized morning briefings to users. Part of a shift toward asynchronous, proactive AI assistance, Pulse generates reports on news, schedules, and preferences. Initially available to Professional Plan subscribers, it aims to expand access gradually.

Google introduces VaultGemma, a 1-billion-parameter large language model trained from scratch with differential privacy (DP), offering strong privacy guarantees while maintaining competitive performance. Based on Gemma 2, it uses DP to prevent memorization of training data, making it suitable for sensitive domains like healthcare and finance. Despite increased computational demands from DP noise, VaultGemma achieves results comparable to non-DP models such as GPT-2 1.5B across multiple benchmark

Adobe enhances creative workflows with AI models like Gemini 2.5 Flash Image and FLUX.1 Kontext in Photoshop, plus Firefly Boards for real-time collaboration and generative editing. New features include presets, text/image editing, and non-destructive AI layers.

Elon Musk's xAI sues OpenAI for allegedly stealing trade secrets and recruiting former employees to gain an unfair competitive edge in AI development.

Oracle plans $15B bond sale for corporate use, cloud expansion, and AI investments amid restructuring and strong fiscal growth.

Emergent Labs raises $23M for AI platform enabling enterprises to build apps via natural language, automating code generation, testing, and cybersecurity features.

NVIDIA partners with Alibaba to integrate AI tools into cloud platform for robotics, autonomous vehicles, and smart spaces, amid Alibaba's $50B+ AI investment and global data center expansion.

Google launches Mixboard, an AI-powered tool for creating mood boards, competing with Pinterest. Users generate visuals via text prompts or templates, with editing powered by Nano Banana. Now in U.S. beta, it supports image merging, adjustments, and text generation, targeting design, decor, and DIY use cases.

Microsoft expands collaboration with Anthropic, integrating its AI models into Copilot alongside OpenAI. Business users can now choose between OpenAI's advanced reasoning model and Anthropic's Claude Opus 4.1 or Sonnet 4 for tasks like research, tool development, and enterprise agents. Opus 4.1 targets complex reasoning and coding, while Sonnet 4 handles routine tasks, data processing, and content creation.

Microsoft updates GitHub Copilot and Azure Migrate to streamline application modernization and cloud migration, featuring AI-driven code fixes, dependency mapping, security checks, and expanded Azure integration. Azure Accelerate offers professional support for workload migration.