Thesis & Problem
Trust in AI has to be engineered, not assumed.
As AI applications and agents take on longer, more consequential work, organisations need ways to keep them consistent, grounded and open to independent checking. That is the work of AI Trust, Risk and Security Management (TRiSM).
The market
A growing category with no dominant vendor
~$2.3B
Estimated AI TRiSM market size, 2024
$7.4B
Forecast market size by 2030
~21.6%
Forecast compound annual growth rate
Figures from Grand View Research, which also describes the market as fragmented with no dominant vendor.
Named problems
Two problems we are focused on
Agent drift
AI agents working across many steps can move away from their original instructions, goals or facts. Small departures add up, and the agent's work stops matching what it was asked to do.
Reasoning divergence
An AI system's reasoning can part ways with the evidence it was given, or with the answer it finally gives. Without an independent check, that gap is hard to see.
A related problem is continuity: most AI applications either forget past conversations or replay ever-growing histories, so context size and cost keep growing.