Sent to 20 readers
Race to AGI Daily Digest - Thursday, July 9, 2026
Share:
TLDR
OpenAI moves closer to human-like interaction with GPT‑Live voice and new GPT‑5.6 Sol models cleared for public use in the US.
China tightens rules on AI ethics, agents, and anthropomorphic chatbots, putting a legal frame around emotional and human-like behavior.
Beijing considers restricting overseas access to advanced domestic AI models, extending the export-control logic from chips to model weights.
The EU delays enforcement of AI Act high-risk rules to 2027–2028, slowing regulatory pressure just as the US and China move ahead.
Alibaba and Baidu rally while Nvidia and Broadcom rise, signaling a market bet on sovereign AI stacks and the chips behind them.
The Full Story
Following Monday’s UN risk talks, Tuesday’s focus on kids and agents, and Wednesday’s GPT‑5.6 approvals, today’s question is sharper: who gets to build human-like AI, and where is it allowed to run?
On the tech side, OpenAI is pushing straight into that “feels human” space with GPT‑Live voice for ChatGPT. It’s a real-time, conversational mode designed to sound and respond more like a person than a tool. You can dig into the launch details here: GPT‑Live voice for ChatGPT ->. In parallel, another slice of the GPT‑5.6 family, the Sol models, just cleared US review for broad public use: GPT‑5.6 Sol public launch approval ->. Together they show the frontier stack getting more capable and more intimate at the same time, all under the safety-approval lane we talked about yesterday. You can zoom out on OpenAI’s trajectory here: OpenAI profile ->.
Now flip to Beijing. China just tightened AI rules around ethics, agents, and anthropomorphic chatbots, explicitly targeting systems that present as emotional or human-like: China’s new rules on ethics and agents ->. That lands right on our “AI Security Stack Becomes the New Front Line” and “Agentic Operating Systems in regulated industries” storylines: if your AI agent feels like a coworker or caregiver, China wants stricter behavioral guardrails.
At the same time, regulators there are weighing limits on overseas access to advanced domestic models from players like Alibaba, ByteDance, Zhipu AI, and DeepSeek: China weighs restricting overseas model access ->. That’s the model-layer twin of chip export controls, and it feeds directly into our China human-like AI clampdown narrative ->.
Europe is taking a different tack: the EU has formally pushed key AI Act high-risk rules out to 2027–2028: EU AI Act delay on high-risk rules ->. So while the US is clearing more frontier models, China is tightening and localizing, and the EU is slowing implementation.
Markets are reading this as a sovereignty story. Alibaba jumped 11%, Baidu almost 5%, while Nvidia and Broadcom climbed and Tesla slipped. Investors look like they’re betting that tightly controlled Chinese platforms and the chips that feed them will win inside those walls, even as Western regulators argue about how human their own models should feel.
Get This Delivered Daily
Join thousands of AI professionals who start their day with Race to AGI.