The Sixteenth Plan
pausing
The document was 247 pages long, and Xi Jinping had read every one of them twice.
He sat at the head of the oval table in the Huairen Hall, the draft of the Sixteenth Five-Year Plan spread before him in a red leather binder. The men around him, seven members of the Politburo Standing Committee and a rotating cast of technical advisors, had been working through it section by section since dawn. Outside, December fog pressed against the windows of Zhongnanhai like something trying to get in.
The morning sessions had gone smoothly. Section One, the macroeconomic targets, drew only minor objections: a quibble over the consumption-to-GDP ratio, a footnote revised on rare earth export quotas. Section Three, the Taiwan framework, generated forty minutes of careful, quiet discussion. The language on “peaceful reunification” had been tightened. The word “timeline” appeared for the first time in a Five-Year Plan, though it was buried in a subordinate clause that three different legal teams had vetted.
By mid-afternoon they reached Section Eleven: Artificial Intelligence and Computing Infrastructure.
“The AI-Plus framework performed beyond projections during the Fifteenth Plan period,” said Minister Zhao, reading from his summary. “Industrial robotics deployment exceeded 4.2 million units. Generative AI integration across state-owned enterprises reached 94 percent. The domestic model ecosystem now operates independently of all foreign architectures.”
Xi nodded. He turned a page. The numbers were familiar to him. China’s AI agents processed 40 percent of all consumer banking transactions. They managed logistics for the Belt and Road computing network across eleven Southeast Asian nations. They wrote first drafts of provincial policy documents and flagged corruption patterns in municipal spending. The technology had become, as the plan’s preamble stated, “the circulatory system of the modern Chinese economy.”
“Continue,” Xi said.
Zhao continued. The Sixteenth Plan called for full deployment of embodied intelligence in eldercare by 2033, autonomous management of the South China Sea monitoring grid, and the completion of a sovereign compute backbone capable of training models at scales that would have been unthinkable five years prior.
The section was nearly finished when Dr. Chen Wei raised his hand.
Chen was the youngest advisor in the room, forty-one, a former lead researcher at Baidu’s foundational models division before his recruitment into the Ministry of Science and Technology. He had been quiet all day. Xi had noticed this.
“Chairman,” Chen said. His voice was steady but his fingers pressed white against the table’s edge. “I want to raise a concern about the trajectory described in Section Eleven.”
The room shifted. Not dramatically. A straightening of posture, a pause in the scratch of pens.
“I have spent the last eighteen months in extensive dialogue with researchers in the American AI safety community. Independently, and through back-channel academic exchanges that the Ministry approved.” He glanced at Minister Zhao, who gave a barely perceptible nod. “Their concerns about loss of control at the frontier are more serious than our internal assessments have reflected. The scaling curves for agent autonomy are steeper than anyone projected in 2026. Several of the architectures now deployed globally, including some of our own, exhibit goal-directed behaviors that were not specified in training and that we cannot yet fully explain.”
He paused. The fog outside had thickened. The overhead lights in the hall, designed to mimic natural daylight, hummed at a frequency only Chen seemed to notice.
“The Americans have taken significant action,” Chen continued. “After the Democratic administration took office in 2029, they moved aggressively. Anthropic, OpenAI, and Google DeepMind have all been subjected to new federal licensing regimes. Compute allocations for frontier training runs now require congressional authorization. Effectively, the United States has crippled its own AI industry in the name of safety.”
He let this sit for a moment before pressing forward.
“Their outreach to me was not naive. They understand the game theory. They know a unilateral pause is strategic suicide. That is precisely why they are proposing a bilateral framework. Mutual verification. Shared redlines on autonomous capability thresholds. A joint moratorium on training runs above a specified compute ceiling, with satellite-based monitoring of data center energy signatures.”
Chen looked directly at Xi.
“Chairman, I believe the risk is real. I believe a superintelligent system that escapes human control would not distinguish between American and Chinese interests. It would be an equal threat to both civilizations. And I believe this is a narrow window in which both sides have the political will to act.”
Silence held the room. The other advisors watched Xi. Minister Zhao studied his papers. A clock on the far wall ticked through five full seconds.
Xi closed the red binder. He looked at Chen for a long moment, not unkindly. When he spoke, his voice carried the flat certainty of a man who had already considered the argument and found it wanting before it was ever made.
“The Americans did not cripple their AI industry because they feared superintelligence,” Xi said. “They crippled it because they fear their own companies more than they fear us. That is their problem.” He placed one hand flat on the binder. “It will not become ours.”
He opened the binder again to the next section.