KPGV Knowledge Partnership for Global AI Vision

Coalition in formation · Founding partners and researchers invited

International AI Research & Technology Coalition

An international coalition for Global AI from India.

KPGV brings together scientists, researchers, universities, governments, computing infrastructure providers and industry partners to advance frontier artificial intelligence, strengthen sovereign AI capabilities, facilitate scientific cooperation across borders and promote AI that can be governed and verified.

Why KPGV

Frontier AI is too consequential to be built in isolation.

The next decade of AI research will shape science, economies and public services. The capacity to do that research, and to judge its results, is not spread evenly. KPGV exists to work on that gap together.

KPGV is designed as a common platform where these four problems are worked on by many institutions at once.

Capability is concentrated

The compute, data, talent and capital needed for frontier research sit with a small number of organisations and places. Many capable researchers and institutions have no route into that work.

Science is fragmented

Research groups, universities and national programmes often work in parallel, with limited shared methods, benchmarks or infrastructure. Good ideas travel slowly.

Trust needs evidence

As AI enters public services, science and industry, governments and institutions need ways to verify what systems do and how they were built, not only to be told.

Sovereignty needs partnership

Countries and institutions want AI capability they can understand, adapt and govern. For most, that is reachable only through cooperation on talent, compute and hardware.

Six strategic pillars

Six pillars for research, capacity and trust.

Each pillar describes what KPGV intends to pursue. The tags show how far along each one is, so that plans are not mistaken for capabilities.

KPGV Frontier AIResearch Compute &Infrastructure Semiconductors& Hardware Global ScientificNetwork Governed &Verifiable AI Sovereign &Collaborative AI
01Founding focus

Frontier AI Research

Joint research on the capabilities, limits and safe behaviour of advanced AI systems, led by member scientists and open to many technical approaches.

  • Foundation and multimodal model research
  • Reasoning, planning and agentic systems
  • AI for scientific discovery
  • Efficient training and inference
  • Understanding capabilities and limits
02Building toward

AI Compute & Infrastructure

Coordinated access to the compute, data and software environments that frontier research depends on, provided by infrastructure partners and shared under agreed terms.

  • Shared access models for member researchers
  • Interoperable software and research environments
  • Secure data collaboration and stewardship
  • Energy-efficient, resilient facilities
  • Capacity supplied by partners, to be agreed
03Long-term ambition

Semiconductors & Hardware

Exploratory research on the chips, memory, interconnects and systems that AI depends on, to understand options for resilient and diverse hardware supply.

  • AI accelerator architectures
  • Memory and interconnect
  • Energy-efficient hardware and software co-design
  • Hardware security and supply-chain verification
  • Roadmapping with the semiconductor research community

Exploratory only. No manufacturing or fabrication commitment is implied.

04Founding focus

Global Scientific Network

A network of researchers and institutions across borders: exchange, joint supervision, shared benchmarks and a community that outlasts any single project.

  • Researcher exchange and visiting positions
  • Joint doctoral and postdoctoral supervision
  • Seminars, workshops and summer schools
  • Common datasets, benchmarks and publications
  • Career pathways for early-career researchers
05Founding focus

Governed & Verifiable AI

Methods and standards that let institutions check what an AI system does, how it was built and whether it meets stated requirements. No single technical approach is presumed.

  • Evaluation and benchmarking methods
  • Provenance, documentation and audit trails
  • Formal and empirical verification techniques
  • Safety, robustness and security testing
  • Governance frameworks and independent review
06Building toward

Sovereign & Collaborative AI

Helping governments and institutions build capability they can understand, adapt and govern, through partnership rather than dependence.

  • Capacity building and training programmes
  • Adaptation to local languages, laws and needs
  • Public-interest applications
  • Dialogue between policy-makers and scientists
  • Respect for data and jurisdiction requirements

How We Work

From shared talent to deployed, verified systems.

KPGV works through four overlapping stages. They repeat as programmes mature, and each is governed by written agreements between the parties involved.

Unite Talent

Bring scientists, engineers and institutions into one coalition through clear membership pathways and shared standards.

In practiceVisiting positions, joint supervision and working groups, as they are established.

Collaborative Research

Form joint programmes around shared questions. Contributors publish and share results under agreed terms for authorship, data and intellectual property.

In practiceA proposed starting agenda is set out in the next section.

Enable Infrastructure

Work with infrastructure providers to arrange access to compute, data and secure research environments for member scientists.

In practiceAccess is arranged through written agreements. Provider agreements are still to be negotiated.

Build & Deploy

Turn validated research into tools, evaluations and systems that members, governments and industry partners can adopt, with verification built in.

In practiceRelease terms are decided for each programme.

Written agreements

Access to resources, data and results is arranged in writing between named parties.

Independent verification

Where a programme makes claims about a system, the evidence should be open to independent review.

No single approach

KPGV does not presume one architecture, one vendor or one method. Approaches are chosen on the evidence.

Research & Development

Working across the frontiers of AI.

KPGV seeks to encourage collaboration across foundational and emerging areas of artificial intelligence.

Foundation models

Pretraining, scaling and model architectures.

General intelligence

Reasoning, planning and advanced capabilities.

Multilingual AI

Language, speech and multimodal intelligence.

Agentic AI

Autonomous systems and intelligent agents.

AI safety

Alignment, robustness and interpretability.

Physical AI

Robotics, autonomous systems and embodied AI.

AI systems

Distributed training, inference and optimisation.

AI hardware Long-term

Accelerators, chip architecture and computing.

Proposed research workflow

Programmes are expected to follow a common path, so that results from different teams can be compared and checked.

  1. QuestionA shared problem set by members.
  2. MethodApproaches proposed and documented.
  3. Shared evaluationCommon tests and benchmarks.
  4. Independent verificationReview by parties outside the team.
  5. ReleasePublication or release as agreed.

International Cooperation

A coalition without borders.

The advancement of artificial intelligence is a shared global opportunity and responsibility. KPGV seeks to facilitate cooperation across countries, disciplines and institutions.

Participation is open to institutions anywhere, subject to applicable law and to the terms agreed with each partner.

  • Governments & sovereign institutions
  • Universities & research centres
  • AI laboratories & technology companies
  • Semiconductor & compute providers
  • Scientific communities
  • Development & research organisations

What each partner brings and gains

Universities and research institutes

Contribute
Researchers, students, laboratories and domain expertise.
Gain
Joint programmes, shared evaluation resources and international collaboration.

Governments and public institutions

Contribute
Policy priorities, public-interest use cases and convening power.
Gain
Technical advice, assurance methods and capacity building.

Computing infrastructure providers

Contribute
Compute, data-centre and cloud capacity, under written agreements.
Gain
A coordinated route to scientific users and research partners.

Industry partners

Contribute
Engineering expertise, hardware and software, and applied problems.
Gain
Verified methods, access to researchers and pre-competitive research.

Researchers and strategic advisors

Contribute
Scientific leadership, independent judgement and introductions.
Gain
A shared platform for cross-border research and a voice in the coalition's direction.

Founding partners

To be announced. Institutions appear here only after they have confirmed in writing.

[Founding partner list — to be confirmed]

Agreements first

Resources, data and results are shared under written agreements between named parties.

Sovereignty respected

Participation does not transfer control of national data, systems or policy.

Law and compliance

Activity follows applicable national law, including export-control and sanctions rules.

Transparent governance

How decisions are made, and by whom, will be published as governance is settled.

Join the Coalition

Five pathways to take part.

Choose the pathway that fits your role. Each opens a short form. Your application is an expression of interest and creates no obligation on either side.

Researchers

For scientists and engineers in universities, institutes and industry laboratories.

  • Contribute to joint research programmes
  • Join working groups on evaluation, systems or governance
  • Take part in exchanges as they are established

Universities

For university leaders, deans and research offices.

  • Explore institutional collaboration
  • Co-develop joint supervision and exchange schemes
  • Help shape partnership terms for universities

Governments

For ministries, agencies and public research bodies.

  • Explore sovereign AI capability building
  • Engage on assurance and governance methods
  • Take part in dialogue between policy-makers and scientists

Industry

For companies working in AI, computing, semiconductors and related fields.

  • Offer compute, hardware or engineering expertise
  • Bring applied problems for joint research
  • Support deployment with verification built in

Strategic advisors

For senior scientists, practitioners and public figures willing to advise on direction.

  • Advise on research direction and governance
  • Open introductions to institutions
  • Review proposals as the coalition forms

KPGV is in formation. Applications are expressions of interest, not offers of membership, and will be handled as the coalition's processes are put in place.

Application

Legal