Limitations

Context

It is important to remember that questions can be asked only in the context of the displayed attack path. Anything not related to the attack path will NOT be answered (for example, who’s Michael Jackson?)

This limitation is being done by using prompt engineering to guide the AI engine to rely only on a specific context (in this case, the attack path context). For anything unrelated, the AI is configured not to respond. This helps the AI to avoid the possibility of generating malicious code.

Random text without context might result in random responses or a request to rephrase the question. For example, in the following screenshot, we typed o and pressed "Ask AI." The AI asked us to rephrase the question.

The same question could lead to different answers if asked twice. The context should still be the same, but we shouldn’t expect the technology to provide the same response every time. There’s a variable in GenAI called temperature which is used to control the diversity or determinism of the AI. High temperatures (closer to 1) provide more creative responses but are less consistent, whereas low temperatures (closer to 0) are more deterministic but less creative. We use a temperature of 0.7, which is considered common in the industry.

Rate Limit

The AI Assistant can generate up to 50 summaries per minute. If the summaries generated are new (that is, not previously viewed and cached by Attack Path Analysis), you may experience a brief unavailability while the AI generates the content.

Quota

The questions quota for the AI Assistant is set to 100 questions per month. You will not be able to ask more questions once the quota is exceeded. If you need to increase the quota or have any questions about the AI Assistant, contact your CSM.