Clairva Q3 2026: Building the Infrastructure for Real-World AI Data
Q3 was the quarter Clairva moved from foundation-building into execution.
We expanded commercial programmes across licensed content, real-world video, multimodal datasets and specialised training requirements, while continuing to build the platform, research capability and institutional supply underneath them.
AI is moving into the physical world, and that changes the data problem. The next generation of models will need far more than text and scraped internet content. They will need to understand how people move, work, interact with objects, navigate environments and behave across different cultures and contexts. Much of that data is proprietary, rights-constrained, poorly structured or does not yet exist in a usable form.
That is the infrastructure problem Clairva is built to solve.
Commercial momentum
During Q3, we expanded our work across egocentric video, real-world activity, scripted content, specialised environments and software repositories.
We now have multiple programmes in evaluation, delivery or commercial discussion with significant AI and technology companies. Some require highly specific physical-world datasets. Others are evaluating large pools of licensed content or new categories of proprietary training data.
The requirements are becoming more sophisticated. Buyers want rights clarity, provenance, precise specifications, consistent quality, traceability and the ability to iterate quickly as model requirements change.
This is increasingly where the value sits.
The platform underneath the data
The Clairva platform is becoming the operating layer behind these programmes.
We are bringing together rights and provenance tracking, consent management, dataset intake, quality control, task orchestration, metadata and delivery workflows into a single system.
At scale, customers need to know where the data came from, what rights attach to it, whether it meets specification, whether it can be used for training and whether the entire chain is auditable. Our objective is to make that process repeatable across customers, content categories and geographies.
The platform is also beginning to create opportunities in adjacent areas. MedTech is one example, where high-quality, rights-aware and carefully governed real-world data has obvious applications. We expect other specialist domains to emerge as the platform becomes capable of handling increasingly complex data environments.
Institutional supply as a strategic asset
We continued to deepen relationships with major content owners, broadcasters, studios, archives and specialist rights holders.
This gives Clairva access to large pools of professionally produced, rights-cleared content while allowing us to structure that supply around specific AI training requirements. Where existing content is insufficient, we can create purpose-built datasets around customer needs.
The strongest AI data infrastructure will need to combine institutional content, proprietary libraries and newly created real-world data into one usable system.
That is the architecture we are building.
From volume to intelligence
Much of the AI data industry still measures value in hours, files and terabytes. We think that is temporary.
The more important question is: which data actually improves the model?
Our research effort is increasingly organised around that problem. The team is developing systems around semantic relevance, task visibility, occlusion, dataset quality and task-video correlation, with the longer-term goal of understanding the relationship between dataset composition and model performance.
Our CTO and Head of AI Research, Sabari Raju, is co-author of work on privacy-preserving training being presented at ECCV 2026. A second paper will also be presented in Florida, extending the research footprint we are building alongside the commercial business.
Research is not a side activity for Clairva. Over time, we want the platform not only to know how to source and govern data, but also to understand which data is most useful for a particular model or task.
That is where the intelligence layer begins.
Building beyond Singapore
Q3 also marked the beginning of a broader geographic push.
Singapore remains our base and an important bridge into Asia, but our customers, partners and market are global. We are now building a stronger presence in both the United States and Europe, where a significant share of frontier AI demand, research and capital sits.
In November, the team will be in the San Francisco Bay Area, meeting AI labs, technology companies, research organisations, partners and investors. The objective is straightforward: move closer to the centre of demand and build relationships that can materially accelerate the next phase of the company.
Europe is equally important, particularly around institutional content, AI research, privacy and rights infrastructure, and we expect our presence there to deepen over the coming quarters.
What Q3 reinforced
Three things became obvious this quarter.
First, the buying cycle for serious AI data programmes is complex. Rights review, compliance, procurement, sample testing and technical evaluation all take time. This complexity increasingly favours companies that can build infrastructure around it.
Second, quality is no longer a simple technical metric. Task visibility, semantic relevance, provenance, rights and model usefulness are becoming just as important as resolution or bitrate.
Third, supply depth matters. The harder the data is to access and govern, the more valuable long-term relationships with rights holders become.
These dynamics favour scale, technology and patience.
What comes next
Our focus for Q4 is direct: convert advanced opportunities into production programmes, deepen institutional supply, strengthen the platform, expand our presence in the US and Europe, explore adjacent verticals such as MedTech, and push the research layer closer to measurable model improvement.
There is a great deal to build, and the market is moving faster than the normal cadence of company-building would suggest. The constraint for us is increasingly not the size of the opportunity, but time: how quickly we can convert demand, expand supply, build the technology, deepen the research and establish Clairva in the markets that matter.
That is a good problem to have.
The next phase of AI will require high-quality, rights-cleared, real-world data at a scale and level of complexity that does not exist today. The companies that can organize that access will become part of the infrastructure of the AI economy.
Clairva is building for that position, and there is far too much to get done to move slowly.