Geopolitics & Sovereign Positioning

The Conference on Neural Information Processing Systems initially announced a policy barring submissions from US-sanctioned institutions including Huawei Technologies and other major Chinese tech groups, prompting the China Computer Federation and other professional bodies to urge Chinese researchers to boycott the event. South China Morning Post The conference organisers subsequently apologised and clarified the ban was more limited than initially indicated. South China Morning Post

The incident reveals how US export controls and sanctions are fragmenting the global AI research ecosystem. Academic conferences that historically operated as neutral scientific forums are now forced to navigate geopolitical restrictions, creating parallel research communities along bloc lines. Chinese institutions are responding not with acquiescence but with collective action to pressure reversal or establish alternative venues.

The Chinese Academy of Sciences unveiled Xiangshan, a high-performance processor achieving 16.5 points per GHz under SPEC2006 benchmarks, representing a significant advance in RISC-V based chips. South China Morning Post Separately, SMIC released an action plan focusing on optimising existing businesses and identifying new growth opportunities, explicitly framing its role as backbone of China's technology self-sufficiency drive. South China Morning Post

RISC-V's open-source architecture allows Chinese firms to develop sophisticated processors without licensing proprietary designs from ARM or using x86 architectures controlled by Intel and AMD. The performance levels now being achieved suggest China is successfully building indigenous chip capabilities that reduce strategic vulnerability to Western export controls, though leading-edge manufacturing processes remain constrained.

A federal judge rejected the Pentagon's attempt to immediately enforce what the court characterised as efforts to cripple Anthropic over the company's restrictions on military use of its AI systems. BBC Defense Secretary Hegseth warned Anthropic to allow the military to use its AI technology without constraints, particularly regarding autonomous weapons and surveillance applications, according to reporting citing unnamed sources. Georgetown CSET

The conflict represents a fundamental tension in US AI strategy between maintaining technology leadership through private sector innovation and ensuring military access to cutting-edge capabilities. Anthropic has positioned itself as prioritising AI safety and ethical constraints, while the Pentagon argues national security requires unfettered access to the most capable systems.

Token consumption on Chinese AI platforms has increased approximately tenfold since late January, with usage doubling every two weeks according to Infinigence CEO Xia Lixue speaking at a state-backed conference. South China Morning Post The surge is attributed to widespread adoption of OpenClaw, an open-source AI agent that has generated nationwide enthusiasm and accelerated practical AI deployment across Chinese enterprises and government entities.

The rapid growth in token consumption indicates both expanding Chinese AI infrastructure capacity and increasing real-world AI application beyond experimental use cases. This organic demand growth creates a self-reinforcing ecosystem where infrastructure investment is justified by actual usage, reducing dependence on Western cloud providers and model access.