Everyone is focused on the tools. That misses the point. FIFA and Lenovo did not just release new software. They changed who gets to interpret the game, and how decisions might be made under pressure.
By David Findlay, Founder of KiqIQ. Connect: dlfindlay
What FIFA and Lenovo Actually Announced
At Lenovo Tech World during CES on January 7, 2026, FIFA and Lenovo introduced three AI systems for the World Cup. The scale matters. Forty eight teams, 104 matches, millions in stadiums, billions watching. This is not a controlled test. It is a global deployment.
The three systems:
Football AI Pro is a generative assistant built on what FIFA calls a Football Language Model. It is trained on decades of match data. Every team gets access. It produces text analysis, video, graphs, and 3D outputs across multiple languages. It cannot be used during live play. Only before and after matches.
AI-enabled 3D player avatars generate full-body volumetric models from scans that take about a second. These feed into semi-automated offside technology, improving measurement accuracy and making VAR graphics more realistic. The system was trialled in 2025.
Next-gen referee view stabilises referee body-cam footage in real time. The goal is simple. Remove motion blur from fast, unstable footage and make it usable for broadcast.
FIFA frames this as democratisation. That is directionally correct. It is also incomplete. This sits inside football data and systems, where access is only one layer of advantage.
Technical Deep-Dive 1: The Football Language Model
A standard language model works on text. This system almost certainly does not.
Training on hundreds of millions of football data points means structured event data. Passes, shots, positional coordinates, outcomes. These are sequences, not sentences. The model has to learn patterns across time and space, not just language. That places it closer to a multimodal foundation model.
The video output exposes a fault line. Is the system retrieving real clips or generating them?
If it retrieves, accuracy can be checked against known footage. If it generates, the burden shifts to the analyst. You now have to verify the output before trusting it. That slows decision making, especially in short turnaround scenarios.
The same dependency runs through the 3D outputs. Visualisations rely on positional accuracy. That links directly to the avatar system and offside tracking. These tools share a pipeline. If one layer is wrong, the rest inherit that error.
Now add deployment reality. Sixteen cities, three countries, different infrastructure and data rules. Latency, bandwidth, and compliance are not uniform. FIFA has not explained how that is handled at system level.
Technical Deep-Dive 2: Offside and the Avatar Pipeline
Semi-automated offside is already in use. It tracks around 29 body points using stadium cameras. The weakness is depth. A 2D projection of a 3D body introduces error, especially at angle.
The avatar system addresses this. A volumetric model captures actual body geometry. The system knows where a limb exists in space, not just where it appears.
That should improve tight decisions.
The open question is update frequency. A single scan at tournament entry creates a static model. Bodies change. Weight shifts, swelling occurs, muscle load fluctuates. If scans are not updated, accuracy drifts as the tournament progresses.
There is also a wider implication for football performance science. Accurate body models support injury analysis and biomechanical profiling. Whether FIFA allows that use is unclear.
Technical Deep-Dive 3: Referee Body-Cam Stabilisation
On paper, this is a broadcast upgrade.
In practice, it is a data upgrade.
The system combines motion data with AI stabilisation to produce clean footage. Once the footage is stable, it becomes machine-readable. That is the shift.
Stable footage allows analysis of proximity, contact, and body orientation. These are the raw inputs for automated foul detection.
FIFA has not announced that capability. It does not need to. The infrastructure now exists.
The trials at recent tournaments provide the dataset required to build that next layer.
The KiqIQ Angle
Equal tools do not equal outcomes
Access is only one variable.
Teams with established analytics environments know how to ask better questions. They know which outputs to trust and how to act on them under time pressure. That knowledge compounds over time.
Smaller federations now have the same tool. They do not have the same experience using it.
The gap does not disappear. It shifts. From data access to interpretation.
What happens after 2026
This tournament is a live test at global scale.
The data generated will refine the model. That has commercial value. It is difficult to see this remaining a universal free tool.
A tiered licensing model is the likely outcome. Clubs with resources gain access. Others do not.
The same inequality reappears in a different place.
Biometric data and control
A full-body scan is biometric data.
Under GDPR and US state laws such as BIPA, that requires explicit consent, defined usage, and limited retention. FIFA has not detailed the consent structure. It has not outlined who can access the data or how long it is stored.
With players from multiple jurisdictions, this is a complex compliance problem, not a footnote.
Referee footage as infrastructure
The rule is simple. Governance should exist before capability.
Stabilised referee footage creates a dataset that can support automated decision systems. Whether or not FIFA intends to build those systems is secondary.
The data already enables it.
IFAB and governing bodies need to define limits now. Waiting until the technology is deployed shifts control away from regulation.
Key Takeaways
- Football AI Pro operates closer to a multimodal system than a standard language model, combining structured match data with generative outputs.
- 3D player avatars improve offside accuracy by replacing 2D projections with real volumetric geometry.
- Referee body-cam stabilisation turns broadcast footage into machine-readable data that could support future automated decisions.
- Equal access to tools shifts competitive advantage toward teams with stronger analytical capability.
- Player scans are biometric data, and the consent and governance framework has not been clearly disclosed.
- The post-tournament commercial model is likely to determine who benefits long term.