Advanced Machine Intelligence

Advanced Machine Intelligence Labs, also known as AMI Labs, is a Paris-based artificial-intelligence research laboratory established in December 2025 and co-founded by Yann LeCun, the former Chief AI Scientist of Meta Platforms. The company is organized around the development of world-model AI systems rather than a conventional large-language-model product strategy. Its stated research direction is to create systems that can represent and understand aspects of the physical world, retain information over long periods, plan and strategize across complicated tasks, and remain secure and controllable in operation. AMI Labs emerged after LeCun left Meta in November 2025. His departure was associated with a criticism of the prevailing concentration of the leading AI companies on scaling large language models. AMI Labs was formed to pursue a broader conception of machine intelligence, in which models would be capable of more persistent memory, physical-world understanding, and structured reasoning about actions and consequences. The laboratory's approach places it within a growing research field commonly described as world models, alongside work by organizations including World Labs, Google DeepMind, OpenAI, and Meta. The company reportedly raised $1.03 billion in a seed financing round involving Bezos Expeditions, Nvidia, Samsung Electronics, and other investors. This funding gives AMI Labs substantial resources for a research-led organization at an early stage, although the available reference material does not identify the complete investor list, ownership structure, or financing terms beyond the reported round size. AMI Labs is not expected to bring a saleable product to market for approximately five years, indicating that its near-term identity is primarily that of a foundational research organization rather than a commercial software vendor. Its initial development model is intended to involve collaboration with companies operating in data-intensive sectors, including manufacturing, biomedicine, and robotics. Such partnerships are designed to provide practical environments and specialized data for testing world-model capabilities. The laboratory's first identified official partner is Nabla, a French startup led by Alexandre LeBrun that develops artificial-intelligence software for physicians, including tools that document patient conversations and generate medical notes. The partnership illustrates AMI Labs' interest in applying foundational research to professional and industrial workflows without implying that AMI Labs itself currently sells a finished medical product. AMI Labs has locations in Paris, New York City, Montreal, and Singapore. These sites give the organization a multi-regional footprint spanning European, North American, and Asian technology ecosystems. Its public positioning emphasizes research depth, long-term machine intelligence, and safety or manageability rather than consumer branding, near-term product distribution, or a specific vertical market. Because the organization was newly established in 2025 and the available material contains limited operational detail, information about its workforce, executive structure, product roadmap, revenue, and commercial customers remains unavailable.

Advanced Machine Intelligence Labs was established in Paris in December 2025 by Yann LeCun, who had previously served as Meta Platforms' Chief AI Scientist. Its formation followed LeCun's departure from Meta in November 2025 and reflected a strategic disagreement with the dominant emphasis among major AI companies on scaling large language models. AMI Labs was created to investigate a broader model of machine intelligence based on systems that can build representations of the physical world, preserve memories over extended periods, plan multi-step activities, and execute complicated tasks while remaining secure and manageable. The laboratory's founding concept places it in the emerging world-model research community. This area seeks to give artificial-intelligence systems richer internal models of environments and events than those generally associated with text-prediction systems alone. AMI Labs is not the only organization pursuing this direction: World Labs, Google DeepMind, OpenAI, and Meta have also been identified as competitors or parallel researchers in related fields. The available material does not establish that AMI Labs has a distinct commercial product advantage over these organizations, but it does distinguish the laboratory by making world-model research its central institutional purpose. At launch, AMI Labs reportedly raised $1.03 billion in seed financing from investors including Bezos Expeditions, Nvidia, and Samsung Electronics. The round is notable for the scale of its reported capital relative to the laboratory's early stage, but detailed financing terms, ownership percentages, and the full investor roster are not available in the supplied references. The organization is expected to remain research-oriented for approximately five years before producing a saleable product, so its early development is likely to be judged primarily through research progress, partnerships, and technical demonstrations rather than ordinary product sales. AMI Labs intends to develop its systems in cooperation with businesses that generate or depend on large volumes of specialized data. Manufacturing, biomedicine, and robotics are cited as initial areas of interest because they offer complex physical environments and practical tasks in which planning, memory, and world understanding could be tested. Its first official partner is Nabla, a French startup led by Alexandre LeBrun. Nabla develops AI tools for doctors that can document patient conversations and prepare medical notes, providing AMI Labs with an early application context in healthcare. The partnership should be understood as a research and industry collaboration; the available information does not say that AMI Labs has launched a separate healthcare product. The laboratory has locations in Paris, New York City, Montreal, and Singapore. This footprint connects its French base with research and technology communities in North America and Asia. No reliable details are available in the supplied material about the precise responsibilities of these offices, employee numbers, additional executives, revenue, or a completed product portfolio. As a result, AMI Labs remains best characterized as a newly founded, heavily financed, internationally distributed AI research organization whose long-term objective is to develop controllable world-model systems for intelligent machines.

• 2025

  Yann LeCun leaves Meta

  [1]Advanced Machine Intelligence Labsen.wikipedia.org

  Yann LeCun departed Meta in November 2025 amid a broader argument that the industry's concentration on large language models could reach important limits.

• 2025

  AMI Labs is founded

  [1]Advanced Machine Intelligence Labsen.wikipedia.org

  Advanced Machine Intelligence Labs was established in Paris in December 2025 to research world-model AI systems.

• 2025

  Seed financing is reported

  [1]Advanced Machine Intelligence Labsen.wikipedia.org

  The laboratory reportedly raised $1.03 billion in seed funding from a group including Bezos Expeditions, Nvidia, and Samsung Electronics.

• 2025

  Nabla becomes the first identified official partner

  [1]Advanced Machine Intelligence Labsen.wikipedia.org

  AMI Labs began an official collaboration with Nabla, a French company developing AI software for medical documentation.

A research-first AI laboratory pursuing world-model systems with physical-world understanding, persistent memory, complex-task planning, and an emphasis on security and controllability.

• Artificial-intelligence research · 2025

  World-model AI systems

  AMI Labs' core research program is aimed at systems that model aspects of the physical world, maintain long-term memory, and plan or strategize through complicated tasks. The systems are also intended to be secure and manageable. The available material describes this as a research direction rather than a launched commercial product, and the organization is not expected to release a saleable offering for approximately five years.

• 2025

  Adopt a research-first world-model strategyStrategy

  Background

  AMI Labs was formed in the context of industry debate over whether large language models alone can deliver broadly capable machine intelligence.

  Action taken

  The laboratory centered its work on world models designed to understand physical environments, retain long-term memory, plan complex tasks, and remain controllable.

  Aftermath

  The organization positioned itself as a long-horizon research laboratory and is not expected to produce a saleable product for about five years.

  • World Labs — World Labs is pursuing related world-model research in the United States; no specific reaction to AMI Labs is identified.
  • Google DeepMind — Google DeepMind has also worked on related approaches; no specific reaction to AMI Labs is identified.
  • OpenAI — OpenAI has been associated with related world-model research; no specific reaction to AMI Labs is identified.
  • Meta — Meta has pursued related research, although no specific response to the founding of AMI Labs is identified.

  Source

• 2025

  Collaborate with data-intensive industriesOther

  Background

  World-model systems require demanding real-world environments and specialized data for evaluation and development.

  Action taken

  AMI Labs planned to work with companies in manufacturing, biomedicine, and robotics, with Nabla identified as its first official partner.

  Aftermath

  The partnership model linked AMI Labs' foundational research to applied settings, beginning with medical documentation workflows through Nabla.

  Source

1. 1Advanced Machine Intelligence Labsen.wikipedia.org
2. 2AMI Labs official websiteamilabs.xyz