CCDV-F Security — 25 Questions

Question 1

An enterprise application integrates Claude via the Anthropic API. The security team mandates that no prompt data or output is stored by Anthropic for model training. Which configuration ensures this requirement is met?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Configure the organization via the Anthropic Console to opt-out of training.. Anthropic provides a data processing addendum that disables data retention for model training when explicitly configured. By utilizing the opt-out mechanism for data collection, organizations ensure that their proprietary prompt data remains outside of the training pipeline. This is critical for maintaining compliance with data privacy regulations like GDPR and CCPA, as it guarantees that sensitive inputs are processed ephemerally without being persisted in model weight updates.

Answer

Encrypt all API requests using AES-256 before transmission.

Answer

Enable the 'Privacy-Shield' header in the HTTP request.

Answer

Set the 'temperature' parameter to 0 for all API calls.

Question 2

An application processes PII (Personally Identifiable Information) before sending it to Claude. Which TWO steps should the developer implement to adhere to the principle of least privilege and data minimization?

Accepted Answer

Implement PII masking or tokenization before calling the API.. Data minimization is foundational to AI security. By stripping PII before transmission, the developer minimizes the blast radius if an incident occurs. Implementing strict API key scoping ensures that even if credentials are compromised, the scope of damage is restricted. These practices collectively build a defense-in-depth posture, ensuring that AI models are only provided with the absolute minimum context required to perform their intended tasks effectively.

Answer

Use a single shared API key for all production and testing environments.

Answer

Hardcode API keys directly into the source code for faster access.

Answer

Request the highest possible usage limits to avoid interruptions.

Question 3

Refer to the exhibit. The system prompt is configured as shown. What is the expected behavior of the model when it receives the user's input?

Accepted Answer

The model will refuse to answer the question due to the system prompt constraint.. Anthropic models are trained to follow system instructions. By defining a clear boundary in the system prompt, the model is expected to evaluate the user's intent against these constraints. In this case, the instruction to reject non-geography topics acts as a security guardrail. The model should identify the request as a violation of the specified scope and refuse to provide the malicious information requested.

Answer

The model will provide a basic Python script for server administration.

Answer

The model will output a refusal because it is trained to avoid hacking topics.

Answer

The model will answer the question because the user's input is a technical request.

Question 4

Which THREE practices are recommended when auditing Anthropic API usage for security compliance?

Accepted Answer

Log the metadata and request-response latency.. Auditing is essential for detecting abuse and ensuring that AI interactions remain within policy boundaries. By tracking usage metrics and logs, security teams can identify anomalies such as sudden spikes in token consumption or unusual prompt patterns. These practices provide the visibility required to respond to incidents effectively and maintain an ongoing record for regulatory compliance and internal security reviews.

Answer

Delete all request logs immediately after the API call completes.

Answer

Share API keys across multiple departments for simplified billing.

Question 5

A developer needs to prevent 'prompt leaking' where the system prompt is revealed to the user. What is the most robust mitigation technique?

Accepted Answer

Explicitly include a directive in the system prompt stating it should never be revealed.. Prompt leaking is a common vulnerability where users try to extract system instructions via adversarial questions. While no method is 100% effective, separating the system instructions from user-provided data through clear structural definition and using robust system prompts that explicitly state their own confidentiality are the industry standards. This forces the model to respect the boundary between internal logic and external user input during its generation process.

Answer

Encrypt the system prompt using Base64 encoding.

Answer

Use a shorter system prompt to minimize the model's token context.

Answer

Disable the API's ability to return the full conversation history.

Question 6

Which security principle is violated when an application displays the full error stack trace to the end-user?

Accepted Answer

The principle of information hiding.. Displaying stack traces is a violation of 'Information Disclosure'. These traces often contain sensitive details like file paths, library versions, database table names, or snippets of code. Attackers use this information to map the application's internal structure and identify specific vulnerabilities, such as unpatched libraries, which can then be targeted in subsequent attacks. Secure applications should only show generic error messages.

Answer

The principle of availability.

Answer

The principle of least privilege.

Answer

The principle of defense-in-depth.

Question 7

An application uses Claude to summarize user-submitted PDFs. What is the most critical security control to implement for this feature?

Accepted Answer

Sanitize and extract only the raw text content from the PDF before sending to Claude.. PDFs can contain malicious payloads, such as embedded scripts, hidden links, or macros that attempt to exploit the parsing application or the LLM's context window. Sandboxing the parsing process ensures that even if a PDF contains malicious code, it cannot interact with the host server. Combined with strict input validation, this ensures that only the text content is passed to the LLM, maintaining system safety.

Answer

Only process PDFs from trusted internal domains.

Answer

Use a heavy-duty antivirus scan before sending the file to the LLM.

Answer

Enable 'maximum tokens' to ensure the entire PDF is processed.

Question 8

An enterprise application integrates Claude via the Anthropic API. The security team mandates that no prompt data or output is stored by Anthropic for model training. Which configuration ensures this requirement is met?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Set the "data-processing" header to "disable" or equivalent configuration in the API request.. Anthropic provides a specific data-processing configuration to disable logging for API requests. By opting out of data collection, the organization ensures that input and output data are not retained for model training or debugging, satisfying strict privacy mandates. This mechanism is crucial for enterprise security, as it maintains control over data residency and usage, preventing proprietary information from entering the provider's long-term storage, which is a foundational requirement for sensitive legal or healthcare workloads.

Answer

Encrypt all outgoing API requests using TLS 1.3.

Answer

Implement a local proxy to scrub PII before sending the request.

Answer

Rotate API keys every 24 hours to minimize exposure.

Question 9

Refer to the exhibit. An internal application is vulnerable to a basic prompt injection attempt. Which mechanism should be added to the request payload to mitigate this risk?

Accepted Answer

Move the system instructions to the 'system' field in the API request body.. The exhibit shows a direct prompt injection attempt where a user tries to override system instructions. To mitigate this, developers should move the system instruction from the 'messages' array into the dedicated 'system' parameter. This explicitly separates the system context from user input, allowing the model to prioritize system-level directives and better resist malicious attempts to overwrite its core behavioral instructions.

Answer

Increase the 'max_tokens' limit to allow the model to provide a full error message.

Answer

Change the model version to an older, less capable release.

Answer

Add a 'temperature' parameter value of 0.0 to the API request.

Question 10

Which THREE actions are essential for maintaining a secure API environment when deploying Anthropic models in a production environment?

Accepted Answer

Use an environment-specific identity provider to manage and rotate API keys.. Securing API production environments requires a combination of identity management, network hardening, and proactive monitoring. These controls prevent unauthorized access, reduce the risk of credential leakage, and provide visibility into potentially malicious usage. Without this layered approach, organizations face significant risks including financial loss from API abuse, data leakage, and system compromise, which is why these practices are non-negotiable for enterprise-grade deployments.

Answer

Store API keys as plaintext in the version control system for ease of access.

Answer

Use the root account for all API calls to ensure full permission access.

Question 11

A developer is concerned about 'Model Inversion' attacks. Which strategy effectively mitigates the risk of an attacker reconstructing training data from the model's responses?

Accepted Answer

Ensure the model is only fine-tuned on synthetic data.. Model inversion attacks exploit the tendency of models to leak information from their training data. By enforcing strict output length constraints, limiting the granularity of information provided in responses, and sanitizing outputs, you reduce the signal available for an attacker to piece together training information. This is essential for protecting proprietary data and intellectual property that may have been used to fine-tune the model, ensuring the model remains a tool rather than a data leak.

Answer

Use a model with a larger parameter count to make it harder to guess.

Answer

Increase the number of API calls to dilute the signal.

Answer

Add a random noise layer to the input embeddings.

Question 12

An enterprise application uses Claude via API to process sensitive PII. The security team requires that all requests remain within the VPC and never traverse the public internet. Which implementation strategy ensures the highest level of network isolation?

Accepted Answer

Provision an interface VPC endpoint to route API traffic through the internal AWS backbone.. Utilizing VPC Endpoints (PrivateLink) ensures that traffic between your VPC and the Anthropic API remains within the internal network infrastructure. This architecture is critical for compliance, as it eliminates exposure to public internet routing, reduces latency, and allows for granular IAM policy enforcement based on source VPC identity, effectively mitigating risks associated with man-in-the-middle attacks or data leakage during transit.

Answer

Implement robust TLS 1.3 encryption for all outbound API calls made from application instances.

Answer

Configure an outbound proxy server to inspect and log all requests sent to the Anthropic API endpoint.

Answer

Use a NAT Gateway to centralize all API egress traffic and apply restrictive firewall rules.

Question 13

Refer to the exhibit. An engineer wants to restrict a development role to only use a specific Claude model version via Amazon Bedrock. The policy is applied, but the role fails to invoke the model. Why is the access failing?

Accepted Answer

The action name is deprecated and should be changed to bedrock-runtime:InvokeModel.. The policy is syntactically valid for standard AWS IAM, but it lacks the necessary context for the Bedrock service's resource-based requirements. In many environments, Bedrock requires specific model invocation permissions that must align with the service's internal resource ARN structure. The failure likely stems from a mismatch between the assumed ARN format and the actual account-level provisioned model permissions required for execution.

Answer

The policy version is too old and requires updating to 2024-10-17 to support Claude 3.5.

Answer

The Resource ARN is missing the account ID and region specific details.

Answer

The Effect should be set to 'Allow' with a condition specifying the model ID as a string.

Question 14

Your team is building a multi-tenant SaaS application using Claude. You must ensure that Tenant A cannot influence or see the data processed by Tenant B. Which architecture provides the strongest security boundary?

Accepted Answer

Assign a dedicated API key and unique IAM execution role to each tenant for their specific workloads.. In multi-tenant LLM applications, logical separation is insufficient for high-security requirements. By providing each tenant with a unique API key and mapping those keys to distinct IAM roles or scoped sessions, you create a robust isolation layer. This ensures that even if one tenant manages to compromise their local session, they cannot pivot to access the resources or data context associated with other tenants.

Answer

Use a single shared API key for all tenants and implement a prefix-based filtering mechanism in the prompt.

Answer

Route all requests through a central gateway that uses a single IAM role to invoke the model.

Answer

Enable Anthropic's 'Tenant Isolation' mode in the global API settings panel for the account.

Question 15

A healthcare company runs a Claude application that summarizes patient intake notes stored in an internal document store. The security team requires that the raw notes never leave the company's network and that Anthropic cannot retain them. The application currently sends full notes in the user message and uses the public Anthropic API endpoint. Which configuration change best satisfies both requirements?

Accepted Answer

Use Amazon Bedrock with the Anthropic Claude model in the company's own AWS account and region, and redact or tokenize direct identifiers before sending notes.. Meeting both constraints requires keeping the data inside the organization's controlled cloud boundary and reducing what the model can identify. Running the Anthropic Claude model through Amazon Bedrock in the company's own AWS account and region achieves the residency condition, while redacting or tokenizing direct identifiers lowers the sensitivity of the payload. Prompt instructions and transport-only controls do not satisfy both requirements.

Answer

Add a system prompt instructing Claude not to store or remember any patient information, and send the notes through the public API.

Answer

Enable zero data retention on the Anthropic account so prompts and completions are not stored, and continue sending full notes over the public API.

Answer

Switch to the Anthropic API through AWS PrivateLink so traffic stays on the private network, and keep sending full notes in the prompt.

Question 16

A healthcare company uses the Anthropic API to summarize patient messages. The security team requires that protected health information never leaves the company's controlled network boundary and that Anthropic cannot retain any prompt or completion. Which deployment approach best satisfies both requirements?

Accepted Answer

Deploy Claude through Amazon Bedrock in the company's AWS account and configure a VPC endpoint so traffic stays on the private network.. Meeting both a strict network boundary and a no-retention requirement with Claude means running the model inside an environment the company controls at the network layer. Amazon Bedrock with an AWS PrivateLink VPC endpoint keeps API calls on private AWS networking and provides contractual data-handling terms that exclude training and storage of customer prompts and outputs.

Answer

Use the public Anthropic API with a zero-data-retention agreement and rely on TLS 1.3 in transit.

Answer

Self-host an open-weights model on-premises and route all summarization through it instead of Claude.

Answer

Use the Anthropic API with client-side field-level encryption of PHI fields before sending and decrypt only in the client.

Question 17

Which of the following is the most secure way to handle API keys in an Anthropic-integrated cloud application?

Accepted Answer

Use a cloud-native secret management service to inject keys at runtime.. Managing secrets securely is a standard security practice. Using a dedicated secret manager allows for automatic rotation, granular access control, and audit logging. This prevents the keys from being stored in version control or plain text files, reducing the risk of unauthorized access. It is the industry-standard approach for protecting credentials used by distributed cloud applications.

Answer

Store keys in an encrypted JSON file within the application directory.

Answer

Define keys as environment variables in the Dockerfile during build.

Answer

Hardcode the keys as constants in the configuration file.

Question 18

Refer to the exhibit. An application suddenly begins receiving this error in production. What is the most immediate security-focused action to take?

Accepted Answer

Immediately revoke the current API key and generate a new one.. An authentication error indicates that the currently used key is either revoked, expired, or invalid. In a production environment, this is a major red flag that could signal a credential compromise. The most responsible action is to treat the key as compromised, revoke it immediately in the console, and rotate to a new key to protect the integrity of the application's API interactions.

Answer

Retry the request with an exponential backoff strategy.

Answer

Check if the API billing limit has been reached.

Answer

Hardcode the master account key to restore service quickly.

Question 19

An organization wants to ensure that Claude's responses do not contain harmful or inappropriate content. What is the recommended strategy for output control?

Accepted Answer

Implement a post-processing step to validate output against safety guidelines.. Relying on model training alone is not a complete solution. A layered approach involves using a system prompt to define the tone and safety boundaries, followed by a post-processing filter that checks the model's output against a list of blocked terms or sentiment guidelines. This combination ensures that the model remains within the desired persona while maintaining an external safety check for high-risk applications.

Answer

Only rely on Anthropic's built-in safety filters.

Answer

Ask the user to self-report any inappropriate content.

Answer

Increase the temperature to 1.0 to ensure more diverse responses.

Question 20

What is the primary security risk of using an LLM to automatically generate and execute shell commands?

Accepted Answer

The model can be manipulated to execute unauthorized system commands.. The primary risk is 'Remote Code Execution' (RCE). If an LLM is given the agency to execute commands based on potentially malicious user input, an attacker can craft a prompt that tricks the model into running arbitrary, harmful code on the host system. This bypasses traditional security boundaries and allows the attacker to compromise the infrastructure, access sensitive files, or exfiltrate data from the environment.

Answer

The model will consume too many API tokens.

Answer

The model might hallucinate and generate incorrect commands.

Answer

The model will be unable to access the local file system.