Explore 2 source-supported AI security cases for 1.4 · Ineffective storage and encryption in Databricks DASF 3.0, with evidence, scope, and defensive analysis.
Risk category
Insecure storage allows unauthorized access to raw data, while ineffective encryption leaves the contents readable if other defenses are bypassed. A successful intrusion or bulk exfiltration can therefore expose the underlying information. The consequences can include data breaches and legal, financial, or reputational harm, especially where storage protections are required by regulation.
Zeabur, a platform for hosting applications and databases, reported an October 14-15 intrusion affecting hosted AI-service credentials. Attackers used leaked internal credentials across cluster and cloud boundaries; a replacement credential was captured in a later backup and subsequently used without authorization. Customers own upstream-key revocation and recovery of deployments they administer; the operator owns internal isolation and backup access.
Why it relates: Plaintext credential storage and secret-bearing backups enlarged the exposure of the AI-hosting environment. Operational secrets are the shared storage risk; no claim that an ML raw dataset or cryptographic primitive was broken.
October operational incident only. Shared mechanisms and AI-hosting layers can justify bounded practical overlap; no inference that an agent, MCP server or particular LLM was deployed. Same-author second pass, not independent validation.
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: Unauthorized access to customer datasets exposes the stored-data confidentiality risk covered by this storage-protection category. No encryption mode, key-management weakness or failed cryptography is identified; the overlap concerns unauthorized access to stored data.
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.