Apple filed a blockbuster lawsuit against OpenAI alleging a coordinated campaign to extract hardware and chip trade secrets through departing employees. On the same day, Anthropic disclosed a $47B annualized run rate — surpassing OpenAI’s $25–33B projection — while OpenAI quietly filed a confidential S-1 targeting a September IPO at up to $850B.
Apple filed a landmark federal lawsuit against OpenAI alleging its hardware chiefs orchestrated systematic theft of Apple trade secrets — a stunning reversal for two companies that partnered publicly in 2024. On the same day, Meta rapidly pulled its controversial Instagram AI image feature after widespread backlash over consent, and Hugging Face’s CEO made the case for why open-source AI matters more than ever.
OpenAI unveiled the GPT-5.6 family (Sol, Terra, Luna) alongside ChatGPT Work — the same day Meta released Muse Spark 1.1 with its first paid API tier. The launches cap a day of turbulence at OpenAI: the copyright lawsuit took a damaging turn after evidence of hidden training-data logs surfaced, and No. 2 executive Fidji Simo stepped down due to illness.
After a 12-day White House-coordinated preview limited to vetted partners, OpenAI’s GPT-5.6 family (Sol, Terra, Luna) is now publicly available alongside the debut of ChatGPT Work. The New York Times coalition simultaneously filed a sanctions motion accusing OpenAI of hiding evidence in the landmark copyright trial, while TypeScript 7.0’s 8–12× build speedups and Ollama’s $88M raise mark a busy day for the developer ecosystem.
An exceptional week for model releases arrives Thursday with OpenAI’s GPT-5.6 Sol, Terra, and Luna going public, while SpaceXAI and Cursor launch Grok 4.5 as early as Wednesday — both targeting Opus-class capability at lower cost. Simultaneously, Microsoft has begun replacing OpenAI and Anthropic models in Excel and Outlook with its own MAI lineup, and China is weighing restrictions on foreign access to its top AI models while DeepSeek moves to build its own inference chips.
Anthropic brings Claude Cowork to mobile and web, revealing that 90% of real-world agentic usage is business and knowledge work, not coding. The same day, Anthropic signed a $19B infrastructure deal with TeraWulf, and security researchers confirmed the first known AI-assisted ransomware attack while Forterra’s self-driving ATVs became the first American autonomous ground vehicles deployed in active combat.
Microsoft opened its fiscal year by eliminating 4,800 roles in commercial sales and Xbox — the latest signal that AI-driven workforce restructuring has moved from prediction to reality. Simultaneously, 169 nations convened in Geneva for the first-ever UN Global AI Governance Dialogue, and OpenAI quietly moved GPT-5.6 into a three-tier preview.
Anthropic’s Claude Science workbench signals AI’s shift from research tool to active scientific participant, while Amazon’s decision to wind down Mechanical Turk marks a symbolic end to human-powered crowdsourcing. Meanwhile, OpenAI’s GPT-5.6 trio remains restricted to a handful of government-vetted organizations as the frontier model race intensifies.
Alibaba moves to ban Claude Code company-wide by July 10, citing alleged hidden tracking of Chinese users — while newly disclosed Pentagon emails reveal Anthropic was blacklisted for refusing to authorize Claude for autonomous weapons. On this July 4th, US-China AI decoupling and military AI ethics are the defining themes of the day.
Anthropic launched Claude Science, an AI workbench for biopharma research paired with an in-house drug discovery program targeting neglected diseases. Meta’s in-training Watermelon model reportedly matches GPT-5.5 on benchmarks, while Mark Zuckerberg admitted internally that AI agent progress has been slower than expected — a candid note from the company that made agents the centerpiece of its 2026 restructuring.