Infinia is all about the process and the workflow, the sources come second. But in 2023 Generative AI came of age.

We have now completed (as of February 2024) our launch of a Federated AI conversation capability, based upon multiple vendor AI APIs.

First off, one thing is very clear to us; that the InfiniaPRO platform is tailor made for the inclusion of AI.

Secondly, we are aware that the output of a LLM query originated in the Infinia platform cannot be taken at face value, think “hallucianations”. Furthermore, simply adding more LLM sources is the equivalent of throwing more haystacks at the needle. Accordingly, the Analyst is driven to identify relevant content from the LLM response and where required seek the evidence to support the O/P from the LLM Query. The Infinia platform accordingly requires the Analyst to review the “evidence” derived from the subsequent OS Searches (after the LLM query response) and weight the evidence in accordance with the classic Admiralty System for the grading of intelligence sources.

Thirdly, our platform supports the Analyst by identifying entities (people, places, events, groups etc) on-the-fly in the LLM response and the Analyst is empowered to determine the relevance or indeed the irrelevance. InfiniaPRO specifically allows the Analyst to capture their thoughts at this point in the process (through their Analyst Commentary) and this addresses the Auditability and Verification of the Analysis in the event of a future need to review the work.

Fourthly, the Open Source query posed to an LLM is no different to queries being posed to other (more historic) Search Engines – it obviously reveals the user’s areas of interest, as has ever been the case since 1998. The sum of a whole sequence of queries would similarly be revealing – this is not for Infinia to manage or de-risk, it is one of Policy. Infinia will however enable the implementation and oversight of Policy. We have built business rules into the Platform that forces the Analyst to be made aware of the associated risks and explicitly accept them or prevent the search. That decision is captured in the Audit Trail.

Finally, ChatGPT (Open AI), Claude and (google) Bard are but three examples of commercially available AI Large Language Models – there are many more and these will grow as businesses like Mistral, LLaMa2 and Jais in the UAE mature. Where we can, we will offer the Analyst the choice to select the best AI capability for the task at hand. We remain LLM agnostic.