OpenAI, DeepMind push US AI coalition as Congress debates kill switch

OpenAI, Anthropic and Google DeepMind CEOs met with President Trump and G7 leaders on June 17 to demand a U.S.‑led coalition on AI safety, a move that comes as Congress debates a “kill switch” bill and Anthropic recently disabled its newest models after export controls were imposed【1】.

High‑level push for international rules

The three CEOs used the G7 forum to argue that the United States should spearhead an “international forum for discussion” that would set globally accepted testing standards and coordinate on cyber, bioterrorism and intelligence risks linked to frontier AI models【1】. Their proposals were echoed by Canada’s prime minister, who signaled support for a U.S.‑led coalition. The meeting coincided with the U.S. government’s recent export‑control action that forced Anthropic to pull access to its Fable 5 and Mythos 5 models, underscoring the urgency of coordinated policy【1】.

OpenAI’s own contribution was a briefing that highlighted the limited‑preview rollout of GPT‑5.5 Cyber to selected cybersecurity teams, a step aimed at demonstrating responsible deployment while the broader regulatory framework remains unsettled【1】.

Competitive backdrop: Google’s expanding user base

While the AI safety dialogue unfolds in Washington, Google’s Gemini chatbot is closing the gap with ChatGPT on the consumer front. Gemini reached 950 million monthly active users in Q2 2026, up from 750 million in February, representing roughly a 27 % increase in a few months【2】. By contrast, OpenAI’s ChatGPT hit the one‑billion‑user milestone in May 2026, according to third‑party estimates【2】. The rapid growth of Gemini underscores the strategic importance of distribution channels that Google can leverage across Search, Android and Workspace, a factor that could shape future policy discussions about market concentration and AI safety.

What to watch

• Congressional action – The Senate Judiciary Committee is slated to hold a hearing on the AI “kill switch” proposal in early August; the outcome will affect export‑control enforcement and industry compliance.
• Further model restrictions – Anthropic’s ongoing negotiations with the Trump administration may lead to additional model shutdowns if export rules tighten.
• Google’s next user milestone – Gemini’s trajectory toward the one‑billion‑user mark could prompt regulatory scrutiny of market dominance and influence the framing of international AI standards.

The convergence of high‑level lobbying, immediate model restrictions, and fierce market competition suggests that the forthcoming policy decisions will not only shape the technical safeguards around AI but also determine which firms can scale their offerings under a unified regulatory regime.