Skip to content

Cloudflare AI governance security checks

AI assistants, agents and model access inside the tenant — what they can read, who can publish them, and how autonomously they act.

On Cloudflare, Black Cat runs 12 checks in this area on every scan. Each one below lists its severity, how to fix it, and the compliance controls it satisfies where a control applies. See what access the Cloudflare connector needs.

Checks (12)

severity: high AI Gateway Logging Disabled fix difficulty: easy #

Enable request logging on the Cloudflare AI Gateway to capture prompts and responses

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Enable Collect Logs under Settings
  4. Verify logs are appearing in the Logs tab

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: high AI Gateway No Authentication fix difficulty: easy #

Enable authentication on the AI Gateway to restrict who can route inference requests

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Enable Authentication under Settings
  4. Distribute the gateway token to authorized callers

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: high AI Gateway No Rate Limiting fix difficulty: easy #

Configure rate limiting on the AI Gateway to prevent abuse and cost overruns

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Configure Rate Limiting with an appropriate interval and limit
  4. Choose a rate limiting technique (fixed or sliding window)

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: medium AI Gateway ZDR Disabled fix difficulty: easy #

Enable Zero Data Retention (ZDR) on the AI Gateway to prevent Cloudflare from storing inference data

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Enable Zero Data Retention under Privacy settings
  4. Note that ZDR may limit some analytics features

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: medium AI Gateway No Guardrails fix difficulty: medium #

Configure content guardrails on the AI Gateway to filter harmful or inappropriate content

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Configure Guardrails under the Content Filtering section
  4. Test guardrails with sample prompts to verify they work as expected

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: medium AI Gateway No Log Export fix difficulty: medium #

Configure Logpush for the AI Gateway to export logs to an external SIEM or storage

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Enable Logpush and configure a destination (S3, R2, Splunk, etc.)
  4. Verify logs are arriving at the destination

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: medium AI Gateway Log Overflow Silent fix difficulty: easy #

Change the AI Gateway log management strategy from DELETE_OLDEST to STOP_INSERTING to prevent silent log loss

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Change Log Management Strategy to STOP_INSERTING
  4. Monitor log storage and increase limits or enable Logpush to avoid hitting the cap

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: medium AI Gateway Aggressive Retry Without Rate Limit fix difficulty: easy #

Reduce retry attempts or add rate limiting to prevent amplified abuse on the AI Gateway

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway and select the gateway
  3. Either reduce Retry Max Attempts to below 4 or configure Rate Limiting
  4. Both controls should be used together for defense in depth

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: low Worker AI Binding Ungoverned fix difficulty: medium #

Review the Worker source code to verify it routes AI calls through an AI Gateway

  1. Identify the Worker script name from the finding
  2. Review the Worker source code for env.AI.run() calls
  3. Check if the gateway option is passed to route through AI Gateway
  4. If not using a gateway, consider updating the code to use one for logging and rate limiting

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: low AI Gateway Account-Wide No Authentication fix difficulty: easy #

Enable authentication on at least one AI Gateway in the account

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > AI Gateway
  3. Review all gateways and enable Authentication on production gateways
  4. Ensure at least one gateway requires authentication

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: low Fine-Tuned Model Present fix difficulty: easy #

Review account fine-tuned models for proprietary training data exposure

  1. Log in to the Cloudflare Dashboard
  2. Navigate to AI > Models and review fine-tuned models
  3. Assess whether training data contains sensitive or proprietary information
  4. Apply access controls or remove models that should not be shared

Vendor docs ↗

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

severity: high Cloudflare AI Gateway Agent High Risk Unblocked fix difficulty: easy #

Block or restrict high-risk AI Gateway agents with scores above 70

  1. Navigate to AI Agent Governance in SSPM
  2. Review the agent risk factors
  3. Block the agent until risks are mitigated

Satisfies: NIS2 Directive NIS2-21.d DORA (SaaS Security) DORA-28.4

More Cloudflare checks

The information on this page is provided for general informational purposes and is believed to be accurate as of its most recent update. Product names, logos, and trademarks are the property of their respective owners and are used for identification purposes only; their use does not imply any affiliation with or endorsement by those owners. Descriptions of third-party applications and of compliance frameworks are based on publicly available documentation and may change over time.

See these checks run on your stack

Start a free 14-day trial — no credit card required.

Start Free Trial