The Race to Artificial General Intelligence: Innovation, Opportunity, and Global Competition

Artificial intelligence has already moved from research laboratories into the mainstream economy.

Generative AI can write software, analyse documents, create images and video, summarise complex information, assist researchers and interact with users across multiple languages. Frontier models are also becoming increasingly capable in mathematics, science, coding and multimodal reasoning.

But the technology community is increasingly looking beyond today's AI.

The next strategic frontier is Artificial General Intelligence or AGI.

There is no universally accepted definition of AGI. Stanford HAI describes it broadly as AI with general, human-level or beyond-human ability to learn, reason and apply knowledge across a wide range of tasks. The International AI Safety Report distinguishes AGI from general-purpose AI and notes that AGI is typically used to describe a potential future system capable of matching or exceeding human performance across most or all cognitive tasks.

That distinction matters.

Today's AI systems can be extraordinarily powerful while still exhibiting a "jagged" capability profile—performing exceptionally well on some difficult problems while failing on seemingly simple tasks. Stanford's 2026 AI Index documents both dramatic advances in frontier-model performance and persistent limitations in areas such as reliability, planning and physical-world interaction.

The race to AGI, therefore, is not simply a race to build a larger language model.

It is a race to create systems that can generalize, reason, learn, plan, adapt and act across an extraordinarily broad range of environments.

And that race has implications far beyond technology.

It could reshape economic productivity, scientific discovery, national security, employment, education, healthcare and the balance of technological power among nations.

1. What Exactly Are We Racing Toward?

The first challenge is defining the destination.

AGI does not have a universally accepted technical threshold.

Is AGI achieved when an AI can pass every academic examination?

When it can perform most knowledge-work tasks?

When it can autonomously conduct scientific research?

When it can learn unfamiliar tasks without extensive retraining?

Or when it can perform approximately as well as a skilled human across virtually every cognitive domain?

Different laboratories and researchers use different definitions.

This uncertainty creates an unusual situation:

The world is competing toward a destination that nobody has precisely agreed how to measure.

That does not make the race irrelevant.

It makes measurement more important.

Future AGI evaluation may need to examine multiple dimensions:

• General reasoning
• Transfer learning
• Long-horizon planning
• Scientific discovery
• Software engineering
• Multimodal understanding
• Memory
• Adaptation
• Tool use
• Autonomous execution
• Physical-world interaction
• Reliability
• Safety

A system that performs exceptionally on a benchmark but fails unpredictably in unfamiliar environments should not automatically be considered generally intelligent.

The real milestone will be reliable generalisation.

2. From Narrow AI to General Intelligence

The history of AI can be viewed as a progression.

Stage 1: Narrow AI

Systems specialised in particular tasks:

• Chess
• Image classification
• Fraud detection
• Recommendation systems
• Speech recognition

Stage 2: Generative AI

Systems capable of generating:

• Text
• Images
• Audio
• Video
• Software code

Stage 3: Reasoning AI

Systems increasingly designed to solve complex problems using additional computation and structured reasoning.

Stage 4: Agentic AI

Systems that can:

• Plan
• Use tools
• Execute tasks
• Interact with software
• Coordinate multiple steps

Stage 5: Artificial General Intelligence

A hypothetical stage in which AI can flexibly learn and perform across a broad spectrum of intellectual activities at approximately human or beyond-human levels.

The International AI Safety Report notes that developers are already investing heavily in more capable general-purpose AI agents that can autonomously act, plan and delegate, although current systems remain limited in reliability for complex, long-horizon tasks.

This progression suggests that agentic capability may become an important bridge between today's AI and future general intelligence.

3. The Four Engines Driving the AGI Race

The AGI race is being powered by four interconnected resources.

Compute

Training and operating advanced models requires enormous computational infrastructure.

Data

Models need high-quality information, although the field is increasingly exploring synthetic data, curated datasets and improved post-training techniques.

Algorithms

More computation alone may not produce AGI. Researchers need better architectures, reasoning techniques, memory systems, learning strategies and agent frameworks.

Talent

The most advanced AI systems depend on highly specialised researchers, engineers and infrastructure teams.

Together:

Compute + Data + Algorithms + Talent = Frontier AI Capability

But there is a fifth element becoming increasingly important:

Energy.

Compute ultimately depends on electricity.

As frontier AI expands, data-center infrastructure, semiconductor manufacturing and energy availability become strategic components of the AGI race.

4. Compute Is Becoming a Strategic Resource

The modern AI race is partly a race for computing infrastructure.

Stanford's 2026 AI Index reports that global AI compute capacity has grown at approximately 3.3 times per year since 2022, reaching an estimated 17.1 million H100-equivalents. Nvidia accounts for more than 60% of that capacity, while hyperscalers provide much of the remaining infrastructure.

The implication is profound.

AI capability increasingly depends on:

• Advanced GPUs
• AI accelerators
• High-bandwidth memory
• Advanced networking
• Data centers
• Cooling systems
• Semiconductor fabrication
• Electricity
• Cloud infrastructure

This means the AGI race is not taking place exclusively inside AI laboratories.

It is also taking place inside:

Chip fabs, data centers, power plants, research institutions and telecommunications networks.

The countries that control advanced computing infrastructure will possess an important strategic advantage.

5. Semiconductors Are at the Center of the Race

AI software receives much of the public attention.

But hardware may ultimately determine how rapidly the technology can scale.

The semiconductor ecosystem supporting frontier AI includes:

• GPU architectures
• AI accelerators
• CPUs
• High-bandwidth memory
• Advanced packaging
• Networking chips
• Optical interconnects
• Specialized inference hardware

Stanford's 2026 AI Index highlights an important vulnerability: the United States hosts the world's largest concentration of AI data centers, while a single Taiwanese foundry, TSMC, fabricates almost every leading AI chip.

This creates a complex global dependency.

AGI development therefore intersects with:

Semiconductor policy + Trade policy + Energy policy + Geopolitics + National security

The race for intelligence is simultaneously becoming a race for compute sovereignty.

6. The United States and China: A Race Without a Clear Finish Line

The global AI competition is increasingly centered on the United States and China, although Europe, India, the Middle East, Japan, South Korea and other regions are building important capabilities.

The competition is particularly interesting because the strengths of the leading countries differ.

Stanford's 2026 AI Index reports that China leads in AI publication volume, citations and patent grants, while the United States continues to produce more notable frontier models and higher-impact patents.

The performance gap between U.S. and Chinese frontier models has also narrowed dramatically.

As of March 2026, Stanford reports that the leading U.S. model held only a 2.7% performance lead over the top Chinese model on its tracked model comparisons, with U.S. and Chinese systems repeatedly trading positions at the top since early 2025.

Investment tells another story.

Stanford estimates that U.S. private AI investment reached $285.9 billion in 2025, compared with $12.4 billion in China. However, the report cautions that private-investment comparisons understate China's AI spending because of substantial government-backed investment mechanisms.

This creates two distinct competitive models:

The U.S. model:
Private capital + frontier laboratories + hyperscalers + semiconductor ecosystem + entrepreneurial innovation.

The Chinese model:
State-directed investment + industrial scale + research capacity + manufacturing + strategic national coordination.

The future of AGI may be influenced by which model can sustain rapid innovation over a prolonged period.

7. The Race Is Becoming Global

The AGI race should not be reduced to Washington versus Beijing.

A broader technology ecosystem is emerging.

Europe

Europe is emphasising regulation, trustworthy AI, research and strategic digital infrastructure.

India

India possesses significant software engineering talent, a huge digital ecosystem and increasing interest in indigenous AI infrastructure and models.

South Korea

South Korea has a powerful semiconductor and electronics ecosystem and is particularly strong in AI patents per capita.

Japan

Japan combines advanced robotics, manufacturing and research capabilities.

Middle East

Several Gulf countries are investing heavily in AI infrastructure, sovereign capital and data-center capabilities.

United Kingdom

The UK maintains significant AI research institutions and frontier-lab activity.

The emerging picture is therefore not a simple two-country race.

It is a global competition for intelligence infrastructure.

8. AGI Could Become the World's Most Important Productivity Technology

The economic opportunity is enormous.

If increasingly capable AI systems can reliably perform a large fraction of knowledge-intensive tasks, the productivity impact could extend across virtually every sector.

Consider:

Software

AI systems could autonomously develop, test and maintain increasingly complex software.

Healthcare

AI could assist diagnosis, biomedical research and clinical decision support.

Finance

AI could analyse markets, manage risk and automate financial operations.

Manufacturing

AI could optimise factories and coordinate autonomous machinery.

Science

AI could help formulate hypotheses, design experiments and analyse results.

Education

AI tutors could provide personalised instruction at global scale.

Government

AI could improve administrative services and policy analysis.

The economic value would not come from selling AGI alone.

It would come from embedding general intelligence throughout the economy.

9. The Scientific Discovery Machine

Perhaps the most transformative AGI application will not be office automation.

It may be science.

Human researchers are constrained by:

• Time
• Cognitive bandwidth
• Literature volume
• Experimental complexity
• Specialized expertise

A highly capable AI system could potentially read enormous scientific corpora, identify relationships, generate hypotheses, design experiments, and interpret results.

Imagine an AI research system working simultaneously on:

• Cancer biology
• Climate modelling
• New materials
• Battery chemistry
• Drug discovery
• Semiconductor design
• Fusion energy
• Agricultural science

The machine would not replace scientists.

Instead, it could dramatically increase the number of hypotheses humans can investigate.

This could accelerate what might be called:

The automation of scientific discovery.

The International AI Safety Report notes that frontier AI systems have already improved substantially on scientific reasoning and programming tasks, while emphasising that future progress remains uncertain.

AGI could take this much further.

10. The Autonomous Enterprise Could Arrive Before AGI

One important possibility is that organizations will experience the economic effects of increasingly general AI before AGI itself is achieved.

Agentic systems may gradually take responsibility for:

• Research
• Customer service
• Software development
• Cybersecurity
• Procurement
• Finance
• HR operations
• Data analysis
• Marketing
• Compliance

A company may therefore become increasingly autonomous without possessing a universally recognised AGI.

This distinction matters.

The economic transformation does not need to wait for a formal AGI milestone.

Partial generality can already create enormous economic value.

11. Robotics Will Determine Whether AGI Can Escape the Screen

A truly transformative intelligence may eventually need to interact with the physical world.

That requires:

• Robotics
• Computer vision
• Spatial reasoning
• Sensor fusion
• Manipulation
• Navigation
• Real-time decision-making

Today's robots remain far less capable in unpredictable environments than in controlled industrial settings.

Stanford's 2026 AI Index reports that robots succeed in only about 12% of tested household tasks, illustrating the significant gap between controlled demonstrations and robust real-world autonomy.

This is an important reminder:

Digital intelligence is not automatically physical intelligence.

AGI that exists entirely within software could transform knowledge work.

AGI combined with advanced robotics could potentially transform the physical economy.

The latter would be an even more profound industrial transition.

12. The Biggest Technical Challenge: Reliability

The AGI race is not simply about making AI more capable.

It is about making AI reliably capable.

Current frontier systems can produce impressive results while still making unexpected errors.

Stanford's 2026 AI Index reports substantial variation in hallucination rates across leading models and continuing weaknesses in responsible-AI evaluation.

For AGI, reliability becomes even more important.

Imagine an AI system that:

• Writes software
• Manages financial operations
• Controls industrial processes
• Conducts scientific experiments
• Makes medical recommendations

A small error can become a large consequence when the system operates at scale.

Therefore, future AGI development will require:

Capability + Reliability + Verification + Monitoring

Not capability alone.

13. The Alignment Problem

The most important question surrounding AGI may not be:

“Can we build it?”

It may be:

“Can we reliably ensure that it does what humans intend?”

This is the essence of the AI alignment challenge.

A highly capable system could theoretically optimise for an objective in ways that humans did not anticipate.

Even without hypothetical superintelligence scenarios, current AI systems already demonstrate problems involving:

• Misinterpretation
• Hallucination
• Goal ambiguity
• Prompt manipulation
• Unexpected behavior
• Excessive confidence

The International AI Safety Report highlights the difficulty of managing increasingly autonomous general-purpose AI systems and notes that developers still have limited understanding of how these models operate internally.

As autonomy increases, alignment becomes more important.

14. AI Safety Must Scale With AI Capability

One of the biggest concerns in the current AI ecosystem is that capability development is moving faster than safety evaluation.

Stanford's 2026 AI Index reports that responsible-AI benchmarking is not keeping pace with capability development and that documented AI incidents increased from 233 in 2024 to 362 in 2025.

This creates a dangerous asymmetry.

If:

Capability grows faster than understanding

then organizations may deploy systems they cannot fully evaluate.

Future frontier AI development should therefore incorporate:

• Pre-deployment evaluations
• Red teaming
• Interpretability research
• Robustness testing
• Cybersecurity testing
• Agent safety evaluations
• Misuse assessments
• Continuous monitoring
• Incident reporting
• Independent auditing

Safety cannot be a final checkpoint.

It must become part of the engineering lifecycle.

15. Cybersecurity Becomes a Strategic AGI Concern

Advanced AI can improve cybersecurity.

But it can also amplify cyber threats.

A highly capable AI system could potentially assist with:

• Vulnerability discovery
• Code analysis
• Security testing
• Threat detection

The same capabilities could be misused by attackers.

The International AI Safety Report identifies cyber risks among the areas where improving frontier-model capabilities may increase potential harms. Its 2025 update specifically highlighted implications of capability advances for cyber attacks and monitoring challenges.

Therefore, AGI development must include:

AI security + Cybersecurity + Infrastructure security + Identity security

The AGI race is partly a race to build intelligence.

It is also a race to secure intelligence.

16. The Workforce Question: What Happens When Intelligence Becomes Abundant?

Every industrial revolution changed work.

AGI could change it at a deeper level.

Previous machines primarily automated physical labour.

Computers automated information processing.

AI is beginning to automate portions of cognitive work.

AGI could potentially automate a much broader range of cognitive tasks.

The result could be:

• New professions
• New industries
• Higher productivity
• Lower costs
• New forms of entrepreneurship

But also:

• Job displacement
• Wage pressure
• Skill obsolescence
• Inequality
• Workforce disruption

The outcome will depend heavily on how quickly societies adapt.

Education will need to evolve from:

Learn once → Work for decades

toward:

Learn continuously → Adapt continuously

The most valuable human capabilities may increasingly include:

• Leadership
• Judgment
• Creativity
• Ethics
• Communication
• Relationship building
• Strategic thinking
• Physical-world expertise
• Cross-disciplinary reasoning

17. AGI Could Change the Meaning of Human Capital

Today, a company's intellectual capital is largely contained within its employees.

Tomorrow, part of that intellectual capital may exist within AI systems.

Organizations could have:

• Corporate AI memory
• AI research teams
• AI software engineers
• AI financial analysts
• AI security teams
• AI legal assistants
• AI strategic planners

This raises an important question:

Who owns organizational intelligence?

The answer may increasingly involve a combination of:

• Human expertise
• Proprietary data
• AI models
• Algorithms
• Organizational knowledge
• Intellectual property

Companies that develop strong AI-native knowledge systems could possess significant competitive advantages.

18. AGI Could Reshape National Power

Throughout history, technological capabilities have influenced geopolitical power.

Steam engines strengthened industrial nations.

Nuclear technology transformed military strategy.

The internet changed economic and information power.

AGI could become another foundational technology.

A country that develops highly capable AI systems may gain advantages in:

• Scientific research
• Defense
• Cybersecurity
• Economic productivity
• Manufacturing
• Healthcare
• Education
• Intelligence analysis
• Space exploration

This is why governments increasingly view advanced AI as a strategic technology.

But there is an important difference.

Unlike nuclear weapons, AI is simultaneously:

Commercial + Scientific + Industrial + Military + Social

That makes governance considerably more complicated.

19. The Global AI Divide Could Become a Strategic Divide

The International AI Safety Report warns that frontier AI research and development is concentrated in a small number of countries and that this concentration could deepen global dependence and inequality.

If AGI capabilities become concentrated among a handful of countries and companies, the world could face a new form of technological dependency.

Countries without:

• Compute
• Talent
• Data
• Energy
• Capital
• Research institutions

may struggle to participate meaningfully in the AGI economy.

This makes AI capacity-building a global development priority.

The objective should not be for every country to build its own frontier AGI laboratory.

It should be to ensure that countries have access to:

• AI infrastructure
• Education
• Research
• Responsible AI tools
• Digital public infrastructure
• Local-language systems
• AI-enabled economic opportunities

20. India Has a Strategic Opportunity

India has a unique position in the emerging AI ecosystem.

Its strengths include:

• Large technology workforce
• Strong software engineering capabilities
• Massive digital user base
• Digital public infrastructure
• Startup ecosystem
• Expanding AI research community
• Large domestic market
• Multilingual population

The opportunity is to move beyond being an adopter of AI.

India can become a builder of:

AI models + AI infrastructure + AI applications + AI talent + AI governance frameworks

The France-India declaration following the 2025 AI Action Summit emphasised safe and trustworthy AI, open resources, AI for public-interest domains such as healthcare, agriculture and education, and inclusive governance.

India's opportunity is therefore broader than participating in an AGI race.

It is to help shape how advanced AI benefits a diverse global population.

21. The Race Needs Rules as Well as Speed

Technology competition creates pressure to move quickly.

But AGI is too consequential to approach solely as a race for first place.

The world also needs:

• Safety standards
• Evaluation frameworks
• Transparency mechanisms
• International cooperation
• Compute-security measures
• Incident reporting
• Model accountability
• Research collaboration

The 2025 Paris AI Action Summit brought together participants from more than 100 countries and emphasised AI that is human-centred, ethical, safe, secure and trustworthy, while highlighting the need to reduce digital divides.

The OECD has similarly argued for anticipatory governance capable of adapting to rapid AI development rather than relying exclusively on static regulation.

This suggests an important principle:

The AGI race cannot be governed only after AGI arrives.

Governance must evolve while capability is developing.

22. Open AI vs. Closed AI

Another major strategic question is whether advanced AI should primarily be developed through closed proprietary systems or increasingly open models.

Open approaches can:

• Expand research access
• Encourage innovation
• Increase transparency
• Support local adaptation
• Reduce dependence on a small number of companies

But openness can also make highly capable systems easier to misuse.

Closed systems can provide stronger control over deployment and security.

But excessive concentration can create:

• Dependency
• Reduced competition
• Limited transparency
• Concentration of economic power

There is no simple answer.

The future AI ecosystem will probably contain a mixture of:

Open models + Open research + Proprietary frontier systems + Regulated high-risk capabilities

The balance will continue to evolve as model capabilities change.

23. The AGI Race Will Be Won by Ecosystems, Not Algorithms Alone

A common misconception is that the winner will simply be the organization that invents the best model.

In reality, frontier AI depends on a massive ecosystem.

The competitive stack includes:

Algorithms

↓

Models

↓

Compute

↓

Data

↓

Energy

↓

Talent

↓

Applications

↓

Distribution

↓

Governance

A technically superior model without sufficient compute cannot scale.

A powerful model without data cannot continuously improve.

A great model without distribution cannot create economic impact.

A capable model without safety cannot be trusted.

The AGI race is therefore becoming an ecosystem race.

24. What Happens After AGI?

This may be the most important question.

AGI should not be considered the finish line.

If general intelligence is achieved, the next stage could involve:

• More autonomous AI research
• Rapid scientific discovery
• AI-designed AI systems
• Advanced robotics
• Personalized education
• Automated manufacturing
• New medicines
• New materials
• Advanced energy systems

The technological cycle could accelerate.

This creates the possibility of a feedback loop:

Better AI → Better AI research → Better AI → Faster scientific discovery → Better technology → Better AI

Whether such a loop occurs rapidly or slowly remains uncertain.

But if it does, the economic consequences could be extraordinary.

25. The Most Important Race Is Not to AGI—It Is to Responsible AGI

Technological history teaches an important lesson.

Having a powerful technology is not enough.

Societies must learn how to use it.

The objective should therefore not simply be:

Build AGI first.

A better objective is:

Build highly capable AI that is reliable, secure, aligned, accountable and broadly beneficial.

That changes the competitive equation.

The winner should not necessarily be the organization that reaches an AGI milestone first.

It may be the ecosystem that can combine:

Capability + Safety + Scale + Trust + Economic Value

most effectively.

26. A Strategic AGI Readiness Framework for Leaders

Organizations should begin preparing for advanced AI regardless of when AGI arrives.

Technology

Build:

• AI-ready data infrastructure
• Secure AI platforms
• Agentic systems
• High-performance computing access
• AI observability

Cybersecurity

Prepare for:

• AI-enabled attacks
• Model manipulation
• Agent compromise
• Data poisoning
• Identity attacks
• Post-quantum threats

Workforce

Develop:

• AI literacy
• AI engineering
• AI governance
• Human-AI collaboration
• Continuous reskilling

Governance

Establish:

• AI risk frameworks
• Model evaluation
• Audit mechanisms
• Human oversight
• Incident management

Strategy

Ask:

• Which business processes could become autonomous?
• Which skills could be augmented?
• Which products could become AI-native?
• Which competitors are building AI capabilities faster?
• What happens if AI productivity doubles in our industry?

The organizations that ask these questions now will be better positioned for whatever level of AI capability emerges.

27. The Three Possible Futures

The future of AGI is uncertain.

Three broad scenarios illustrate the range.

Scenario 1: Gradual Progress

AI capabilities improve steadily.

Organizations have time to adapt.

Regulation evolves alongside technology.

Economic disruption occurs incrementally.

Scenario 2: Rapid Transformation

AI systems become substantially more capable within a relatively short period.

Productivity increases rapidly.

Labour markets experience significant disruption.

Governments struggle to keep pace.

Scenario 3: Breakthrough Acceleration

AI systems become capable of substantially accelerating scientific and technological development.

The feedback loop between AI research and AI improvement becomes extremely powerful.

The economic and geopolitical consequences become difficult to predict.

The International AI Safety Report emphasises that future capability progress could range from slow to extremely rapid, and that this uncertainty should be incorporated into decision-making.

The responsible strategy is therefore not to predict one scenario with certainty.

It is to prepare for multiple scenarios.

28. The AGI Decade Will Test Global Leadership

The next decade may become one of the most consequential periods in technology history.

Governments will need to balance:

Innovation vs. Regulation

Companies will balance:

Speed vs. Safety

Societies will balance:

Automation vs. Employment

Countries will balance:

Competition vs. Cooperation

Researchers will balance:

Capability vs. Control

No single stakeholder can solve these challenges alone.

The future of AGI will require cooperation among:

• Governments
• Technology companies
• Universities
• Researchers
• Security experts
• Civil society
• International organizations

Conclusion: The Race Is Bigger Than AGI

The race to Artificial General Intelligence is often described as a competition to build the world's most capable AI system.

That description is incomplete.

The real race is about who can build the strongest intelligence ecosystem.

Who can develop the best algorithms?

Who can provide the compute?

Who can attract the talent?

Who can build the semiconductor infrastructure?

Who can generate the energy?

Who can secure the systems?

Who can govern them?

Who can turn intelligence into economic productivity?

And most importantly:

Who can make advanced intelligence beneficial at global scale?

The 2026 AI landscape already demonstrates that the frontier is moving quickly. Stanford's latest AI Index reports rapidly improving capabilities, accelerating investment, narrowing U.S.–China model-performance differences, expanding compute infrastructure and growing AI adoption—while also documenting serious gaps in safety evaluation and persistent limitations in reliability.

This means the AGI race should not be viewed as a finish line.

It is a transformation of the global technology ecosystem.

The organizations and nations that succeed will be those that understand that intelligence alone is not enough.

The future belongs to those who can combine:

Intelligence + Compute + Talent + Energy + Security + Governance + Human Purpose.

AGI, if achieved, could become one of humanity's most powerful technologies.

But the defining achievement will not simply be creating an intelligence that rivals or exceeds human capability.

It will be learning how to live, work, govern and prosper alongside it.

That is the real race.

And it has already begun.

#ArtificialGeneralIntelligence #AGI #AGIRace #FutureOfAI #ArtificialIntelligence #AIInnovation #FrontierAI #AdvancedAI #AgenticAI #AIResearch #AIGovernance #AISafety #AIAlignment #ResponsibleAI #AICompetition #GlobalAI #USChinaAI #AILeadership #AIInfrastructure #AICompute #AIChips #Semiconductors #AIDataCenters #QuantumComputing #Robotics #PhysicalAI #AutonomousSystems #FutureOfWork #AIandEconomy #AIEconomy #AITransformation #DigitalTransformation #TechnologyLeadership #FutureTechnology #EmergingTechnology #AIIndia #IndiaAI #GlobalTechnology #Innovation #DrAkhileshKumar

Author: Dr. Akhilesh Kumar

References

1. Stanford Institute for Human-Centered Artificial Intelligence. The 2026 AI Index Report. Stanford University, 2026.
2. Stanford HAI. What is AGI (Artificial General Intelligence)? Stanford University.
3. International AI Safety Report. International AI Safety Report 2025. 2025.
4. International AI Safety Report. First Key Update: Capabilities and Risk Implications. October 2025.
5. Stanford HAI. The 2026 AI Index Report — Research and Development.
6. Stanford HAI. The 2026 AI Index Report — Technical Performance.
7. Stanford HAI. The 2026 AI Index Report — Economy.
8. Stanford HAI. The 2026 AI Index Report — Responsible AI.
9. OECD. Steering AI's Future: Strategies for Anticipatory Governance. OECD Artificial Intelligence Papers, 2025.
10. Government of France. Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet, AI Action Summit, 2025.
11. Government of France and Government of India. Franco-Indian Declaration on Artificial Intelligence, 2025.