Apple has filed a 41-page lawsuit against OpenAI in the U.S. District Court for the Northern District of California, alleging that former Apple employees who joined OpenAI stole confidential product information. The complaint includes extraordinary claims about a former engineer exploiting an unknown authentication bug to siphon sensitive files weeks after leaving the company.
Meta has removed a feature that allowed users to modify photos from public Instagram accounts using AI, following widespread user backlash. The feature, rolled out earlier in the week as part of Meta's Muse Image generator suite, was described by the company as having "missed the mark" and is now unavailable.
OpenAI has officially sunset its AI-powered Atlas browser less than a year after launch, folding its agentic browsing capabilities into the ChatGPT desktop app and a new Chrome extension. The move signals OpenAI's conclusion that AI should be a feature within existing tools rather than a standalone browser destination.
Google DeepMind announced major new capabilities for Managed Agents in the Gemini API, including background execution for async interactions, remote Model Context Protocol server integration, custom function calling alongside sandbox tools, and network credential refresh — turning managed agents into reliable production-ready workers.
Microsoft announced a new operating business called Microsoft Frontier Company, dedicated to delivering enterprise AI deployments using its existing AI tools. Backed by a $2.5 billion investment and 6,000 industry and engineering experts, the venture aims to be the largest outcome-driven engineering organization in the industry. Early partnerships include the London Stock Exchange Group, Unilever, and Land O'Lakes.
Following weeks of negotiations with the Trump administration, Anthropic has restored global access to Claude Fable 5, its consumer-facing Mythos-class AI model. The model was suspended in June after the US government imposed export controls citing potential jailbreak vulnerabilities. Anthropic has implemented improved safety classifiers and proposed a shared industry framework for scoring jailbreak severity.
Google has expanded Gemini Spark to macOS, enabling the AI agent to work with local files and compete with Claude Desktop and Copilot. New integrations include Google Tasks, Google Keep, Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals. Custom MCP support allows users to tailor Spark to their specific needs.
Privacy-focused AI platform Venice AI has raised $65M in Series A funding at a $1B valuation, led by Dragonfly with Coinbase Ventures participating. The company serves over 3 million active users, processes 1.7M API calls daily, and generates $70M+ in annualized revenue. CEO Erik Voorhees positions Venice as a neutral AI platform offering access to 200+ models with end-to-end encryption.
Anthropic released Claude Sonnet 5, a model designed to close the gap between Sonnet and Opus-class performance in agentic tasks. It can plan, use browsers and terminals autonomously, and match Opus 4.8 on some benchmarks at a fraction of the cost — available across all plans starting June 30, 2026.
OpenAI announced its newest frontier AI model family, GPT-5.6, with three variants: Sol for complex reasoning, Terra for high-volume business tasks, and Luna for fast everyday work. Initially available to ~20 vetted organizations following US government coordination, the models represent a shift from version-numbering to capability-tier naming. Sol and Terra set new benchmark highs.