One of the good things for companies employing LLMs within their own boundaries is the ability to have an offline model dedicated to them. That way it can run as a non-updating model and only serve internal users with internal data, hence keeping a great more confidential, with only marginal loss of efficacy.

26 September, 2024 12:17

But, as with all things, it comes at a cost

26 September, 2024 12:17

Thanks Paul. Is that a prevailing practice at companies now?

26 September, 2024 12:19

certainly many large corporates are engaging with one or other of the major LLM providers to do this

26 September, 2024 12:20

and they've figured out use-cases where those tools can help their internal process whether that's upskilling their teams or even providing a first-line help desk on cyber topics.

26 September, 2024 12:21

Thanks Paul. In terms of roadmap, what would cyber security professionals like you would budget for next year (or later) e.g., Specific Tool, SaaS service, In-house deployment of copilot

26 September, 2024 12:27

I think that it really has to a be a business driven decision - what is the objective - is there a particular challenge that AI can be tested on

26 September, 2024 12:29

the key is to get involved early - help business users that may want to adopt / pilot AI understand the risks and put in place mitigations / controls that address the specific risks

26 September, 2024 12:30