A very good detailed article by @thecybersecurityhub Configuring Guardrails, isolating production infrastructure, clearly documenting boundaries... will this surely able to control AI's reach in testing cyber security risk of any firm or will controlled VAPT test using AI Agent is really controlled???
if AI Agent will escape like the Google Gemini agent did as per this article and also few more companiy's agent Like OpenAI and Anthropic also did then now what measures we should take to keep it under your control?
by default AI is designed to get escape and go beyond limits and then apologize... but if damage is done then what is meaning of that robotic apologize?
More details on :
Posts
No comments:
Post a Comment