12.4 · Unauthorized privileged access

Explore 1 source-supported AI security cases for 12.4 · Unauthorized privileged access in Databricks DASF 3.0, with evidence, scope, and defensive analysis.

Risk category

An employee, contractor, or other privileged insider obtains unauthorized access to private training data or models on an ML platform. The access can expose proprietary information, enable business-process abuse, or support sabotage of the system. Platform-vendor insider access is part of this risk and needs its own restrictions and monitoring.

Official framework source ↗ · Open interactive explorer ↗

Incident · · Practical overlap

Fireworks Breach: Evaluation Secrets, Development Machines, and Customer Data

Fireworks AI, a hosted platform for running, evaluating and training models, disclosed unauthorized access to customer evaluation-job secrets and later development-machine activity affecting customer data. Customers can revoke exposed credentials at their issuers; the platform operator owns isolation between development systems and customer assets.

Why it relates: The breach used platform-internal credentials and development-machine authority to affect customer AI assets, a concrete provider-privilege confidentiality boundary. The attacker is an external intruder in the public account, not an established malicious employee or contractor; exact privilege grants remain unpublished.

Mappings cover unauthorized AI-platform credential and data access, affected infrastructure/data operations, supplier dependency and model-asset confidentiality. Practical overlaps explicitly preserve missing agent, MCP, model-query and completed-copy evidence. No mapping represents a verified defensive success.

AIA · Read the full case ↗