Thank you, Brian Levine, for posting. What we have learnt from our work developing AI-enabled cybersecurity audit and assurance tools for is that…

← Back to Blog

Thank you, Brian Levine, for posting. What we have learnt from our work developing AI-enabled cybersecurity audit and assurance tools for is that you have to build in ontological guardrails and keep AI working within boundaries. Continually reinforcing to check and recheck its reasoning before reaching conclusions, then getting AI to recheck those conclusions. We don’t just do this to keep AI from straying and hallucinating; we also do it to focus AI on resolving a specific control. There are several techniques we use to keep it from wandering. But above all, you have to keep AI in a box, not let it out. Ever wonder why the US president revoked the use of Anthropic Fable and Mythos?

AI usage in audit and assurance offers many advantages, but AI also has to be trustworthy, and that only comes when it can resolve audit and assurance outcomes as well as a human. AI has to complement humans and not replace them.

Aegis Labs

#artificalintelligence #aiontology

View Original Post
https://www.linkedin.com/feed/update/urn%3Ali%3Ashare%3A7473022996249575424