China's AI Policy Enters 'Crossroads Era' Following DeepSeek-R1's Early 2025 Breakthrough

A recent overview from the Carnegie Endowment for International Peace highlights a pivotal shift in China's artificial intelligence (AI) policy, entering what authors Scott R. Singer and Matt Sheehan term the "Crossroads Era." This new phase, triggered by the emergence of the advanced DeepSeek-R1 model in early 2025, forces Beijing to navigate a complex balance between fostering technological innovation and asserting centralized state control. Adam Thierer, a technology policy expert, underscored the report's significance, tweeting, "for those following the US-China AI race, this is a really good @CarnegieEndow overview of recent Chinese AI policy developments from @Scott_R_Singer & @mattsheehan88."

The Carnegie analysis traces China's AI policy through distinct periods: the ambitious "Go-Go Era" (2017-early 2020) of rapid development and minimal regulation, followed by the "Crackdown Era" (2020-late 2022) focused on ideological control. This led to the "Catch-Up Era" (late 2022-early 2025), prompted by the perceived gap with Western models like ChatGPT and domestic economic pressures, which saw a recalibration towards pro-development policies.

DeepSeek-R1's performance, comparable to leading Western models, has reportedly restored a sense of technological confidence within the Chinese Communist Party. This newfound strength, however, coincides with persistent economic fragility, marked by a nearly 50% year-on-year decline in venture capital funding for Chinese AI startups in Q1 2025, according to the report. This creates an unprecedented policy dilemma for Beijing.

In response, the government is pushing its "AI+" initiative, aiming to integrate AI into traditional economic sectors and critical infrastructure. Simultaneously, there's a renewed emphasis on control, with new "Measures for Labeling of AI-Generated Synthetic Content" adopted in March 2025, requiring both explicit and implicit labeling of AI-generated media. Reports also indicate intensified oversight of leading AI firms like DeepSeek, including restrictions on talent mobility and investor access.

The ongoing US export controls on advanced semiconductors further complicate China's AI ambitions, potentially delaying the development of future models like DeepSeek-R2 due to limited access to high-end chips. Beijing faces the critical choice of prioritizing control, which could stifle innovation, or maintaining the pragmatic flexibility that propelled its recent AI advancements, ultimately shaping its trajectory in the global AI race.