Event details
A key difference I could think of - ChatGPT doesn't have a native integration with Sentinel/Defender to run and test the queries built. Even if someone builds/has a custom integration to do so with ChatGPT, the data would leave your tenant (to OpenAI) and opens up a whole new can of worms from security perspective.
In my opinion - any AI platform can generate KQL queries, analyze incidents, anomalies, etc. But if you care about your data, you'll need to keep it in your tenant and that's where an integrated ecosystem comes into play.
All things considered, ChatGPT or other AI solutions often creates more accurate queries when it comes to complicated tasks. Security Copilot is good for pretty basic queries. Not all the time you need to share the data externally, just the hypothesis or idea, and 3rd party AI is always more accurate than Copilot in my experience from last few months.
Don't get me wrong - I've had similar observations too. A friend of mine built a custom tool (https://github.com/jguimera/TeisecAgent) with personalized logic to reduce hallucinations. But the time and effort was countless. Its a fine balance between using 3rd party tools vs 1st party tool making an ecosystem vs building your own tool to get the benefits of both. Its a hard comparison 🤷♂️