OpenAI began a limited preview of its GPT-5.6 series — Sol as flagship, Terra as balanced, Luna as fast and cheap — restricted to a small group of trusted partners whose participation has been shared with the government, at the administration's request.

The White House had asked OpenAI to stagger the release over safety concerns, reportedly citing the new model's capabilities as on par with Anthropic's Mythos. OpenAI says Sol ships with its most robust safety stack to date and claims better coding-workflow performance than Claude Mythos 5 while using roughly a third of the output tokens; Terra is pitched at GPT-5.5-class performance for half the price. OpenAI framed the restriction as a short-term step, explicitly saying it does not believe this kind of government access process should become the long-term default.

Jonas Adler and Alexander Pritzel — both key contributors to Google's Gemini models — are leaving Google for Anthropic, according to Bloomberg, the latest in a run of senior departures from Google DeepMind.

They follow Nobel laureate John Jumper, who left for Anthropic on June 19, and earlier exits including Noam Shazeer to OpenAI. Fortune reported the departures are raising doubts about whether DeepMind can stay at the forefront of AI development, and TechCrunch noted the timing: with OpenAI and Anthropic racing toward IPOs, equity offers are a powerful recruiting lever against a company whose AI work sits inside a giant ad business.

Anthropic introduced Claude Tag, an always-on Claude that joins Slack as a team member: grant it selected channels plus access to chosen tools, data and codebases, then anyone in the channel can @Claude a task while they work on something else.

Claude builds persistent context from the channels it sits in and can plan tasks for the future; Anthropic describes it as an evolution of Claude Code that makes the model more proactive and better suited to whole teams. In beta for Claude Team and Enterprise customers, it is already deeply dogfooded: Anthropic says 65% of its product team's code is now written by its internal version of Claude Tag, and the pattern is spreading to support tickets, product metrics and root-cause analysis.

Open-weight lab Reflection AI agreed to pay SpaceX $150 million a month from July 1, 2026 through 2029 — roughly $6.3 billion if it runs the full term — for immediate access to Nvidia GB300 chips and supporting hardware at SpaceX's Colossus 2 data center near Memphis.

Reflection, founded in 2024 by two former Google DeepMind researchers, called it one of the largest announced infrastructure commitments to open AI models, and framed the capacity as runway to build the world's best open models at scale. It is the third big tenant for Colossus after Anthropic and Google, as SpaceX becomes a frontier-AI landlord. The same day, inference cloud Groq confirmed $650M in growth capital led by Disruptive and Infinitum, tied to scaling toward 200 MW by 2027.

Google DeepMind and independent film studio A24 announced a multi-project R&D partnership that anchors DeepMind research directly inside the creative process, with A24 filmmakers helping shape new tooling and workflows and giving researchers hands-on feedback.

Google also made an investment in A24 — reported around $75 million, with an AI research lab at the studio — though neither company put a number in the joint announcement. The stated initial focus is bridging cutting-edge technology and next-generation entertainment, with specific technical outputs and creative milestones left to evolve as researchers and filmmakers work side by side.