Advai Platform for AI Testing and Monitoring

Advai Ltd is a London-based AI assurance company co-founded by CEO David Sully, CTO Chris Jefferson and chief research scientist Damian Ruck. Funded since 2020 through UK Defence and Security Accelerator projects and work with Dstl on adversarial attacks against computer vision, it built one of the UK's first dedicated AI assurance capabilities under the motto of breaking AI to make it better.

The Advai Platform has two modules. Testing includes Model Arena, which benchmarks models, vendors and configurations against the buyer's policy, latency and cost constraints; Insights, with test plans tied to governance and compliance requirements and versioned thresholds; and repeatable red teaming that ends in go-live thresholds and an evidence pack. Monitoring tracks logs across AI systems for drift, degradation, policy breaches, security threats, data leakage, cost and tool use. It covers agents, LLMs, computer vision, time series and multimodal systems for insurance, legal, finance, healthcare and defence clients.

• Model Arena: benchmarks models, vendors and configurations against policy, latency and cost constraints.
• Insights: governance-aligned test plans with traceable results and versioned thresholds.
• Red teaming: repeatable adversarial testing for agents, LLMs and computer vision.
• Evidence packs: go-live thresholds and audit-ready records for deployment approval.
• Monitor: log analysis across AI systems for drift, degradation, security and data leakage alerts.
• Managed onboarding: use-case risk assessment, threshold setting and initial test configuration.

Vietnamese insurers and banks can compare LLM vendors on cost, latency and safety before procurement and keep evidence packs for regulators. Public agencies can red team computer vision systems used in surveillance or traffic management. Enterprises running several AI assistants can monitor them centrally for data leakage.

Procurement and risk teams choose models on measured evidence rather than vendor marketing. Each deployment gets explicit go-live thresholds and an evidence pack for sign-off. Continuous monitoring warns early about drift, data leakage and policy breaches after launch. Advai states it is certified to ISO 27001 and Cyber Essentials Plus, which eases supplier due diligence.

Governments and regulated industries are shifting from AI principles to independent testing before deployment, and defence and public agencies increasingly require sovereign assurance of AI they buy. Financial and insurance firms in Southeast Asia face similar supervisory expectations as they deploy GenAI and agents.

Advai Ltd has its registered address at 20-22 Wenlock Road, London. DASA has funded Advai since 2020 and it has worked with Dstl on physical adversarial patches; it contributed to the Turing Institute's AI security framework and demonstrated its technology to the UK Defence Secretary and Chancellor in March 2025. It states ISO 27001 and Cyber Essentials Plus certification.