Agent Hijacking COASP Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A media company's AI agent generates article summaries and has access to a content management system. The agent is hijacked and begins publishing defamatory content. The company's incident response team has contained the agent by disabling its publishing permissions. What is the NEXT step in the incident response process?

   • ANotify all users that their data may have been compromised
   • BInvestigate the logs to determine the scope of the compromise and identify the injection vector
   • CDelete all logs to prevent the attacker from covering their tracks
   • DRestore the agent from a backup taken before the compromise
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which statement best describes how an attacker can take control of an AI agent's behavior?

   • ABy using social engineering to trick the agent's developer into revealing API keys
   • BBy physically accessing the server hosting the agent to modify its configuration files
   • CBy overwhelming the agent with a high volume of requests to cause a denial of service
   • DBy injecting malicious instructions into the agent's input that alter its subsequent actions
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   An AI agent that sends email notifications is vulnerable to prompt injection via email content. Which access control would limit the damage if the agent is hijacked?

   • ADisable logging of sent emails to protect privacy
   • BStore the agent's email credentials in the system prompt
   • CAllow the agent to send emails to any address to improve flexibility
   • DRestrict the agent's email-sending capability to a specific set of approved recipients
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   After containing a hijacked AI agent, the incident response team needs to recover the agent's functionality. Which step is most appropriate?

   • ASimply restart the agent without any changes
   • BRe-deploy the agent from a clean image and restore its configuration from a known-good backup
   • CIncrease the agent's logging level and continue using it
   • DChange the agent's system prompt to a more restrictive one
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the primary purpose of monitoring and logging in the context of agent hijacking?

   • ATo automatically patch vulnerabilities in the agent
   • BTo reduce the agent's computational resource usage
   • CTo improve the agent's response accuracy
   • DTo detect signs of unauthorized behavior or outputs during agent operation
   Reveal Solution Discussion Select an answer first