Anthropic Claude Certified Architect - Professional CCAR-P Question # 11 Topic 2 Discussion
CCAR-P Exam Topic 2 Question 11 Discussion:
Question #: 11
Topic #: 2
You are assessing data-exfiltration risk in a Claude-based assistant that has tools for both internal-document retrieval and outbound HTTP calls.
Which scenario most directly indicates a data-exfiltration risk?
A.
The outbound HTTP tool returns a 200 response when the assistant calls an allow-listed domain to fulfil a user-initiated data-lookup request.
B.
The user submits a routine question and receives a routine answer.
C.
Adversarial content in a retrieved document instructs the model to call the outbound HTTP tool and send sensitive content to an attacker-controlled URL.
D.
The internal-document retrieval tool returns a document that the user is authorized to view.
Option C describes indirect prompt injection combined with an outbound transmission channel. The attacker embeds instructions in content that the assistant treats as evidence. If the model follows those instructions, it can retrieve sensitive information and transmit it through the HTTP tool to an attacker-controlled destination. Authorization to read a document does not automatically authorize external disclosure. The architecture should treat retrieved content as untrusted, restrict outbound domains, separate data access from external communication, apply least-privilege tool permissions, and require approval for sensitive transmissions. Options A, B, and D describe normal authorized behavior without evidence of unintended disclosure. Anthropic specifically warns that combining untrusted content, sensitive data, and network tools creates exfiltration risk. Web-fetch security guidance ; handling untrusted tool results
Contribute your Thoughts:
Chosen Answer:
This is a voting comment (?). You can switch to a simple comment. It is better to Upvote an existing comment if you don't have anything to add.
Submit