Programmes
Scaling Inference Lab
The aim of the Scaling Inference Lab is to rapidly reduce the cost of large AI systems by giving startups a practical environment in which they can test their AI technology at scale.
Backed by £50m, the lab will create 5+ rack-level AI systems, each serving as a structured pilot for a novel technology, allowing it to be validated within a full-system.
By providing a neutral, open testbed where new hardware and systems can be validated in real environments, the lab unlocks:
- A shop window for emerging technologies to demonstrate real performance
- Faster iteration, with new rack-scale systems built and tested every 6 months
- Independent validation that de-risks investment decisions
High Assurance
AI has enormous potential in financial services, healthcare, and other sectors where compliance, correctness and safety are critical, but deployment is held back by regulatory risk, security vulnerabilities, lack of auditability and unpredictable behaviour.
The High Assurance programme aims to remove these barriers by building the technical foundations that make safe, compliant AI deployment achievable in practice, including:
- Verified Outputs — mathematical proof that AI outputs are correct
- Secure Execution — sandboxed environments that isolate AI-generated code
- Multi-model Validation — cross-checking across models to flag potential risks
- Intelligent Filtering — preventing data leakage and policy violations
What is a Programme?
CommonAI programmes bring members together to work collaboratively on a particular area of focus under a robust governance structure.
The key aims of a programme are defined by a programme charter and development priorities outlined by the technical roadmap.
The programme director leads development by chairing:
- the Technical Committee — responsible for the technical decisions required to implement the roadmap,
- and Steering Committee — responsible for managing the programme.
CommonAI members and CommonAI engineers contribute to programme development and work together to deliver items on the roadmap.