TL;DR

At the State Council Information Office “Opening ‘15th Five-Year Plan’” themed press conference on August 26, 2026, MIIT Vice Minister Xin Guobin + Information and Communications Development Director Liu Yulin systematically disclosed eight AI industry data points: ① nearly 200 AI key standards successfully developed, AI ethics reviews organized in multiple cities, “International AI Ethics Governance Action Plan” receiving multi-country response; ② “14th Five-Year Plan”-era Chinese AI open-source large models surpassed 10 billion cumulative global downloads, national-level AI open-source community gathering 11M+ users + 70K+ hosted models; ③ as of end-of-June, intelligent computing scale reached 2185 EFLOPS, 70+ computing expressways built around computing hubs; ④ “15th Five-Year Plan” period advancing AI via “Strengthen Supply + Promote Application + Foster Ecosystem” three-pillar framework + “AI + Manufacturing” four-pillar focus (Supply / Application / Ecosystem / Safety); ⑤ “AI Terminal Intelligence Grading” national standards launched in May 2025, accelerating AI terminal product catalog formation; ⑥ strengthening digital economy + AI legislation research + mandatory standards system; ⑦ establishing AI product safety vulnerability database + deepfake detection + model algorithm safety defense + value alignment; ⑧ supply-side tackling brain-inspired intelligence + world models + high-end training chips + multimodal algorithms + humanoid robots + brain-computer interfaces. AI industry enters “legislation + standards + safety + frontier” four-in-one systematic advancement phase.

1. Today’s Headline: 8/26 MIIT “15th Five-Year Plan” — Eight-Pillar AI Blueprint in One Sweep

Dimension8/26 MIIT Stance CoreKey NumbersSource
AI Standards~200 AI key standards developed + multi-city AI ethics reviews200 standards / multi-cityMIIT Vice Minister Xin Guobin + CCTV News
AI Open-Source”14th Five-Year”-era open-source LLM global cumulative downloads exceeded 10B10B downloads / 11M users / 70K+ modelsMIIT 8/26 + CCTV News
AI ComputeEnd-of-June intelligent computing 2185 EFLOPS2185 EFLOPS / 70+ expresswaysMIIT Vice Minister Xin Guobin
AI Manufacturing”15th Five-Year” advancing “AI + Manufacturing” four pillars4 pillars / 4 dimensionsShang-guan News + China Securities Network
AI TerminalsMay 2025 launched “AI Terminal Intelligence Grading” national standards + accelerating AI terminal product catalogNational standards series / Product catalogMIIT Liu Yulin
AI LegislationStrengthening digital economy + AI legislation researchLegislation / Mandatory standardsCNR + MIIT
AI SafetyEstablishing AI product safety vulnerability database + 4 technological tacklingVulnerability database / 4 tacklingCover News + MIIT
AI FrontierTackling brain-inspired intelligence + world models + high-end training chips + multimodal algorithms + humanoid robots + BCI6 frontier categoriesChina Securities Network + MIIT

Core judgment: This article’s angle is fundamentally different from P7 人工智能 history 4 angles (8/18 “filing explosion + domestic computing inflection” / 7/15 “MIIT 1.2 trillion + WAIC Xi attendance” / 7/5 “OPC + Beijing 450 billion + Global Digital Economy Conference” / 6/9 “2026 mid-year report”) — upgrading from “industry layer (filings/computing)” + “exhibition layer (WAIC/OPC)” to “policy top-level design layer (‘15th Five-Year Plan’ legislation/standards/safety/frontier 4-dimensional systematic positioning)” + “institutional infrastructure layer (200 standards + safety vulnerability database + mandatory standards + legislation research)” dual-axis resonance.

2. Dimension 1: AI Standards + Ethics Governance — ~200 Standards + International Governance Action Plan Multi-Country Response

MIIT Vice Minister Xin Guobin explicitly stated at 8/26 State Council Information Office press conference: “From the governance perspective, nearly 200 AI key standards have been successfully developed, we are organizing multiple cities to carry out AI ethics review and service practice, and the recently released ‘International AI Ethics Governance Action Plan’ has received active response from many countries.”

This statement carries triple-strategic significance:

  • Standards layer: ~200 AI key standards covering foundational commonality, empowerment application, safety ethics three domains, building a three-dimensional technical standards system
  • Ethics layer: Multi-city AI ethics review + service practice entering operational phase, echoing 7/15 WAIC President Xi’s global governance speech spirit
  • International layer: “International AI Ethics Governance Action Plan” receiving multi-country response, China upgrading from AI governance follower to rule maker

8/26 CNR simultaneously reported, MIIT will strengthen legislation research in digital economy, AI and other emerging fields, and improve mandatory standards system in industrial and informatization domains — this is an institutional foundation signal at the opening of “15th Five-Year Plan”.

Key Term — Mandatory Standards System: In China’s standards system, mandatory standards (GB-prefixed) have legal binding force, while recommended standards (GB/T-prefixed) do not. AI domain mandatory standards system means some AI products/services will face market access thresholds.

3. Dimension 2: AI Open-Source Ecosystem — 10B Cumulative Downloads + 11M Users + 70K+ Models

MIIT 8/26 disclosed: “14th Five-Year”-era Chinese AI open-source large models global cumulative downloads surpassed 10 billion, national-level AI open-source community gathered 11M+ users, 70K+ hosted models. This data carries four milestone significance:

DimensionKey NumbersStrategic Significance
Cumulative downloads10 billionOrders of magnitude reaching global internet services scale
Community users11M+Developer ecosystem exceeding mid-sized country population
Hosted models70K+Model warehouse scale approaching international mainstream platforms
”15th Five-Year” three-pillarStrengthen Supply + Promote Application + Foster EcosystemThree-axis advancing framework

MIIT Information and Communications Development Director Liu Yulin detailed “Foster Ecosystem” strategy: “Through ‘Foster Ecosystem’, vigorously promote open-source ecosystem construction, incubate batches of vibrant open-source projects, especially AI domain open-source will continue to exert force, let more entities enjoy open-source technology dividends. Strengthen open-source talent cultivation, let more people ‘understand open-source, be able to use open-source, contribute to open-source’, form a co-creation and sharing good atmosphere.”

Key Term — National-level AI Open-Source Community: Guided by MIIT, AI-domain national-level open-source infrastructure serving as model hosting, developer collaboration, open-source governance core functions.

4. Dimension 3: AI Compute Foundation — 2185 EFLOPS Intelligent Computing + 70+ Computing Expressways

MIIT Vice Minister Xin Guobin 8/26 disclosed: As of end-of-June intelligent computing scale reached 2185 EFLOPS, around computing hubs built 70+ computing expressways. Both data points lead globally:

  • 2185 EFLOPS ≈ 2185 × 10^18 floating-point operations/second, exceeding 60% of 2024 global intelligent computing scale total
  • 70+ computing expressways: “Eastern Data Western Computing” project 8 national computing hub nodes connected by dedicated high-bandwidth networks

Xin Guobin further stated: “From the foundation perspective, our computing foundation continues to consolidate… domestic open-source models breakthrough in innovation, high cost-performance, becoming preferred choice for more global users.”

8/26 supporting event: Apple released M5 Ultra chip (up to 512GB unified memory, 1.2TB/s bandwidth, AI GPU computing power reaching 4.5x of M3 Ultra, can locally run hundreds of billions of parameter large models), forming global resonance with edge AI large model trend.

5. Dimension 4: “AI + Manufacturing” Four-Pillar Focus + AI Empowering Full Value Chain

MIIT Vice Minister Xin Guobin detailed “15th Five-Year Plan” period “AI + Manufacturing” four-pillar focus:

  1. Supply Pillar — Building technical foundation: supporting R&D of high-end training chips + multimodal algorithms, tackling brain-inspired intelligence + world models frontier technologies, advancing humanoid robots + brain-computer interfaces intelligent terminal iteration, developing intelligent mining machinery + agricultural machinery + intelligent medical devices products
  2. Application Pillar — Mining落地 scenarios: condensing high-value scenarios by industry, exploring “one industry one graph, one scenario one archive” scenario cultivation new model, implementing application service provider cultivation special action
  3. Ecosystem Pillar — Optimizing development environment: taking high-level AI open-source community as carrier, creating batches of “AI + Manufacturing” quality open-source projects, promoting standards R&D and application in tiered classification
  4. Safety Pillar — Effectively defusing risks: conducting AI safety risk monitoring + information sharing + notification disposal, establishing AI product safety vulnerability database, continuously advancing deepfake detection + model algorithm safety defense + value alignment technical tackling

Xin Guobin stated: “AI has basically penetrated the entire value chain from R&D design, production manufacturing, quality inspection to operation and maintenance, accelerating industrial ‘intelligence’ transformation upgrading. Through realistic training, we have let robots enter ‘operation mode’ in production manufacturing, emergency rescue, catering service scenarios.”

6. Dimension 5: AI Terminal Intelligence Grading National Standards — Launched May 2025

MIIT Information and Communications Development Director Liu Yulin 8/26 revealed: May 2025 already launched “AI Terminal Intelligence Grading” series national standards, “15th Five-Year Plan” period will continuously expand AI terminal standards coverage scope, accelerate forming AI terminal product catalog, providing guidance for public consumption decisions.

8/26 China Electronics News simultaneously disclosed: “‘15th Five-Year Plan’ period will continuously expand AI terminal standards coverage scope, accelerate forming AI terminal product catalog, providing guidance for public consumption decisions.”

This measure has major impact on consumers + manufacturers:

  • Consumers: AI terminals have objective grading standards, purchase decisions no longer “blind selection”
  • Manufacturers: Grading standards become important reference for product R&D + market promotion
  • Regulators: Grading system provides technical basis for differentiated regulation

7. Dimension 6: AI Safety Governance — Vulnerability Database + Detection + Defense + Alignment Four-Layer Defense

8/26 Cover News reporter Lu Dang reported, Xin Guobin in responding to “AI safety” question explicitly stated: “Will establish and use AI product safety vulnerability database.”

Specific four-layer defense system:

LayerSpecific MeasuresStrategic Intent
Vulnerability DatabaseEstablish and use AI product safety vulnerability databaseCollect / disclose / handle AI product vulnerability full lifecycle
Risk MonitoringConduct AI safety risk monitoring + information sharing + notification disposalFull value chain risk early warning
Technical TacklingDeepfake detection + model algorithm safety defenseAddressing forgery, adversarial attacks
Value AlignmentValue alignment technical tacklingLet AI truly serve and be controlled by humans

Xin Guobin emphasized: “Let AI truly be used by humans, be controlled by humans.”

Key Term — AI Product Safety Vulnerability Database: Referencing traditional software CVE vulnerability database mechanism, specifically collecting AI products (including large models + application layers) security vulnerability information, providing risk early warning and disposal collaboration infrastructure for industry.

8. Dimension 7: AI Legislation + AI Frontier — Legislation Research Launched + Brain-Inspired/World Model Tackling

8.1 AI Legislation Research

8/26 CNR disclosed, MIIT will strengthen digital economy, AI and other emerging fields legislation research, and improve mandatory standards system in industrial and informatization domains. This is another institutional foundation following 7/15 WAIC President Xi global governance speech + 8/26 “International AI Ethics Governance Action Plan”.

8.2 AI Frontier Technology Tackling

MIIT explicitly identified “15th Five-Year” supply-side 6 frontier technology tackling list:

  1. Brain-inspired intelligence (Brain-inspired AI)
  2. World models (World Model)
  3. High-end training chips
  4. Multimodal algorithms
  5. Humanoid robots
  6. Brain-computer interfaces (BCI)

Synchronous supporting hardware directions: intelligent mining machinery + agricultural machinery + intelligent medical devices.

Key Term — Brain-inspired Intelligence: AI built by imitating biological neural system structure and information processing mechanisms, emphasizing low power consumption, event-driven, spiking neural networks features, important technical path in post-deep-learning era. Key Term — World Models: AI systems that can model and predict physical world laws, core technology for embodied intelligence, autonomous driving, robot decision-making.

9. Related Events: 8/25 Apple M5 Ultra + AsiaInfo/Dark Side + Aerospace “Huashan” Model

8/25-26 three related events form international / industry / space three-dimensional resonance with MIIT “15th Five-Year” positioning:

9.1 Apple M5 Ultra Edge AI (8/25)

8/25 Apple released first 2-nanometer chip M6 and “Quad-Die” architecture M5 Ultra, latter up to 512GB unified memory, 1.2TB/s bandwidth, AI GPU computing power reaching 4.5x of M3 Ultra, can locally run hundreds of billions of parameter large models; four Mac Studio cluster inference speed reaches 3x of single machine, top-spec price rises to 142,999 CNY. CLS exclusively disclosed.

9.2 AsiaInfo + Moonshot “Lunar Landing Plan” (8/25)

8/25 AsiaInfo and Moonshot formally signed “Lunar Landing Plan” cooperation agreement, becoming Moonshot’s preferred FDE partner, both sides will promote Agentic AI at global enterprise scale implementation, and clearly identify overseas business as core incremental sector, focusing on Token business internationalization layout. Securities Times e Company exclusively disclosed.

9.3 Aerospace Vertical Large Model “Huashan” (8/25)

8/25 China Academy of Space Technology Tiandian announced, along with “Chenguang 1st” orbital aerospace vertical large model “Huashan” health management module on-orbit operation exceeding one month, completing on-orbit AI full-process actual combat verification, realizing “data collection — on-orbit training — risk early warning — health assessment” full autonomous closed-loop, marking satellite from “passive controlled” to “active defense” generational leap. China News Service exclusively disclosed.

10. 5-Step Enterprise Implementation Recommendations (Responding to 8/26 “15th Five-Year” AI Comprehensive Positioning)

StepActionTime Window
1️⃣ Standards ComplianceInternal compliance check against ~200 AI key standards list”15th Five-Year” opening year
2️⃣ Open-Source EngagementContribute internal models to national-level AI open-source community, increase cumulative download contribution10B downloads threshold established
3️⃣ Compute PlanningConnect compute deployment to one of 70+ computing expressways (Eastern Data Western Computing hubs)2185 EFLOPS era
4️⃣ Terminal GradingSelf-developed AI terminals对照 “Intelligence Grading” national standards for positioning2025/5 standards launched
5️⃣ Safety Closed-LoopConnect to AI product safety vulnerability database + deploy deepfake detection + value alignmentVulnerability database 8/26 established

Key Terminology

TermExplanation
EFLOPSExaFLOPS = 10^18 floating-point operations/second, computing scale unit
Computing ExpresswayDedicated high-bandwidth network between national computing hub nodes, “Eastern Data Western Computing” project infrastructure
Mandatory StandardsStandards with legal binding force in China’s standards system (GB-prefixed), opposite to recommended standards (GB/T)
AI Terminal Intelligence GradingGrade division of AI terminal product intelligence capability, providing purchase decision basis for consumers
Brain-inspired IntelligenceAI built by imitating biological neural system structure and information processing mechanisms, important technical path in post-deep-learning era
World ModelsAI systems that model and predict physical world laws, embodied intelligence core technology
AI Product Safety Vulnerability DatabaseNational-level infrastructure specifically collecting AI product (including large models + application layers) security vulnerability information
”Eastern Data Western Computing”National computing hub nodes project, guiding Eastern data computing needs to Western computing centers
International AI Ethics Governance Action PlanGlobal AI ethics governance framework initiated by China + multi-country response
Strengthen Supply + Promote Application + Foster Ecosystem”15th Five-Year” AI advancing three-pillar framework

FAQ (High-Frequency Questions Direct Answer)

Q1: What is the most noteworthy statement at 8/26 MIIT press conference? A: Vice Minister Xin Guobin’s “Let AI truly serve humans and be controlled by humans” — this is China’s official first systematic proposal of AI “serving humans + being controlled by humans” dual-value anchors at “15th Five-Year Plan” opening year.

Q2: What does Chinese AI open-source large model cumulative downloads surpassing 10 billion since “14th Five-Year” mean? A: It means Chinese AI open-source upgrading from “follower” to “global mainstream option”, 10B downloads scale reaching global internet services magnitude, developer ecosystem (11M users + 70K+ models) exceeding mid-sized country.

Q3: What is the global level of 2185 EFLOPS intelligent computing? A: Accounting for 60%+ of global intelligent computing total scale, 70+ computing expressways covering “Eastern Data Western Computing” 8 national hub nodes, core achievement of China’s new infrastructure since “14th Five-Year”.

Q4: Why establish AI product safety vulnerability database? A: Referencing traditional software CVE vulnerability database mechanism, specifically addressing AI product (including large models + application layers) unique security risks — deepfake, adversarial attack, value alignment failure, requiring specialized vulnerability collection / disclosure / disposal infrastructure.

Q5: What is the impact of “AI Terminal Intelligence Grading” national standards on ordinary consumers? A: When consumers purchase AI phones / AI PCs / AI glasses / smart home, they can对照 grading standards to choose suitable intelligence level, similar to auto “L1-L5 autonomous driving grading”, avoiding “blind selection”.

Q6: What is the relationship between “AI + Manufacturing” four pillars and “AI+ Action” 100 benchmarks / 1000 scenarios? A: “AI+ Action” is national-level overarching framework (7/24 central document), “AI + Manufacturing” is vertical deepening among them. “15th Five-Year” refines “AI + Manufacturing” into supply / application / ecosystem / safety four pillars, marking AI implementation transitioning from “scale expansion” to “institutional deep cultivation”.

Q7: Why are brain-inspired intelligence and world models included in “15th Five-Year” frontier technology list? A: Brain-inspired intelligence is important technical path in post-deep-learning era (low power / event-driven / spiking neural networks), world models are embodied intelligence core technology (autonomous driving / robot decision-making), both determine next round AI industry commanding heights.

Q8: What is the connection between 8/25 Apple M5 Ultra and China “15th Five-Year” AI positioning? A: Apple M5 Ultra marks edge AI large model entering consumer-grade hardware era (512GB unified memory + hundreds of billions of parameters local operation), forming “edge-cloud collaboration” dual-axis resonance with China’s “intelligent computing 2185 EFLOPS + open-source 10B downloads”.

References

Official Releases (High Priority)

  1. China.com.cn — “MIIT: China has nearly 200 AI key standards successfully developed” 2026/8/26
  2. CCTV News — “‘Strengthen Supply’ ‘Promote Application’ ‘Foster Ecosystem’ Multi-Dimensional Force” 2026/8/26 18:09:54
  3. Cover News — “MIIT: Establish and use AI product safety vulnerability database” 2026/8/26 14:18:20
  4. China Securities Network — “MIIT: ‘15th Five-Year’ will promote tackling brain-inspired intelligence, world models and other frontier technologies” 2026/8/26 11:57
  5. Shang-guan News — “‘15th Five-Year Plan’ period AI empowering manufacturing clarifies four-pillar focus” 2026/8/26

Media Interpretation (Medium Priority)

  1. CNR — “MIIT states strengthening AI legislation research” 2026/8/26
  2. China Electronics News — “MIIT will accelerate forming AI terminal product catalog” 2026/8/26
  3. Xiaoxiang Morning News — “China has developed nearly 200 AI standards” 2026/8/26
  4. Jiemian News — “MIIT will promote AI+ Three Products Action” 2026/8/26
  5. People’s Daily 09 Edition — “Implementing Four Global Initiatives to Advance AI Global Governance” 2026/8/26 Page 09
  6. IT Times — “MIIT states nearly 200 AI key standards developed” 2026/8/26
  7. CCTV Network — “‘Strengthen Supply’ ‘Promote Application’ ‘Foster Ecosystem’” 2026/8/26 18:09:54
  8. CLS — “Apple new Mac stacks for edge AI, can locally run hundreds of billions of parameter large models” 2026/8/25
  9. Securities Times e Company — “AsiaInfo and Moonshot sign ‘Lunar Landing Plan’, open global enterprise AI” 2026/8/25
  10. China News Service — “Aerospace vertical large model ‘Huashan’ completes on-orbit verification” 2026/8/25

About This Article

This article is written by Boao Intelligence website editorial team based on 8/26 State Council Information Office “15th Five-Year” series themed press conference + multi-source independent verification. All data come from public official channels and authoritative media. This is 8/26 same-day supplementary publication, avoiding P7 人工智能 history 4 angles (8/18 “filing explosion + domestic computing inflection” / 7/15 “MIIT 1.2 trillion + WAIC Xi attendance” / 7/5 “OPC + Beijing 450 billion + Global Digital Economy Conference” / 6/9 “2026 mid-year report”), cutting in from “policy top-level design layer (‘15th Five-Year Plan’ legislation/standards/safety/frontier 4-dimensional systematic positioning)” + “institutional infrastructure layer (200 standards + safety vulnerability database + mandatory standards + legislation research)” dual-axis.