SSRF Through AI Tool Calls COASP Practice Questions (Page 6)

Questions 26–28

1. 26application · medium

   A security engineer is reviewing an AI application that uses a tool to fetch URLs. The engineer notices that the tool follows redirects. Which of the following is the most important reason to disable redirects or validate the final URL after redirects?

   • ARedirects can cause the AI model to receive unexpected content.
   • BRedirects are not supported by the HTTP client.
   • CRedirects can be used to bypass URL validation and reach internal resources.
   • DRedirects increase the response time and degrade performance.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   What is a potential impact of SSRF in an AI system that has access to internal network resources?

   • AThe attacker can read sensitive data from internal services, such as configuration files or cloud metadata.
   • BThe attacker can cause the AI model to output incorrect answers.
   • CThe attacker can cause a denial of service by overwhelming the AI's GPU resources.
   • DThe attacker can directly modify the AI model's training data.
   Reveal Solution Discussion Select an answer first
3. 28application · medium

   An AI agent is designed to call a weather API. The agent receives a user prompt and extracts a city name, then constructs a URL like 'https://api.weather.com/current?city=' + city. An attacker submits a prompt with a city value of 'test&url=http://169.254.169.254/latest/meta-data/'. What is the primary reason this could lead to SSRF?

   • AThe AI agent uses a weak authentication mechanism.
   • BThe city value is concatenated into the URL without proper encoding, allowing an attacker to inject additional parameters.
   • CThe weather API is vulnerable to SQL injection.
   • DThe AI model may misinterpret the city value and call a different API.
   Reveal Solution Discussion Select an answer first