Fireworks AI disclosed expanded customer data exposure on November 9, 2026, following two security incidents in October. The findings raise concerns about AI infrastructure security, third-party dependencies, and sensitive customer information.
The incidents also prompted responses from connected platforms and customers, including Notion and legal AI company Harvey. For managed service providers (MSPs), they highlight five important security lessons.
1. Understand the Scope of AI Data Exposure
Fireworks AI provides infrastructure for running and training artificial intelligence models. Its investigation identified unauthorized access to account and billing information, resource metadata, and, for certain customers, stored secrets, datasets, and incomplete model weights.
Fireworks stated that payment card numbers were not affected and reported no detected impact to its Inference, Training SDK, or Nexus services.
MSP Takeaway: Review AI vendors’ data storage and access practices to identify potential exposure of sensitive client information.
2. Assess Third-Party Cloud Security Risks
Fireworks initially identified unauthorized use of internal credentials on October 6.
A separate incident involving a third-party cloud provider occurred on October 23. The following day, an unauthorized party accessed Fireworks development machines and executed a workload that resulted in additional customer data exposure.
Fireworks reported detecting the activity and implementing containment measures.
MSP Takeaway: Evaluate AI vendors’ infrastructure dependencies, security controls, and incident response capabilities as part of client risk management.
3. Strengthen AI Credential Protection
The initial October incident involved unauthorized access to environment variables containing customer secrets, including API keys and tokens.
Fireworks subsequently identified possible links to a financially motivated group associated with LLMjacking, which involves unauthorized AI service usage through compromised credentials.
The company revoked and replaced credentials, restricted access, and strengthened monitoring.
MSP Takeaway: Protect API keys, enforce least-privilege access, rotate exposed credentials, and monitor AI services for suspicious activity.
4. Recognize the Impact on AI Customers
The incidents demonstrate how security problems at an AI infrastructure provider can affect connected businesses.
Notion reported that fewer than 1% of its workspaces were affected, including exposure involving some Notion AI Insights chat transcripts. It disconnected Fireworks, disabled Insights, and recommended credential rotation where appropriate.
Separately, The Information reported that Fireworks notified customers, including Harvey, about the incident. A Harvey spokesperson stated that no Harvey customer data was affected.
MSP Takeaway: Monitor vendor notifications, verify client exposure, and distinguish confirmed compromises from precautionary security alerts.
5. Make AI Vendor Oversight an Ongoing Priority
Fireworks implemented additional safeguards, including credential replacement, network isolation, tighter access controls, and enhanced monitoring.
The company also temporarily disabled certain dataset upload and download capabilities while continuing its investigation with Palo Alto Networks Unit 42.
These developments reinforce the importance of evaluating vendor security throughout the business relationship, not just during onboarding.
MSP Takeaway: Regularly assess vendor security practices, review incident disclosures, and maintain contingency plans for critical AI services.
What This Means for MSPs
The Fireworks AI incidents demonstrate how compromised credentials and third-party infrastructure can create risks across interconnected AI services.
For MSPs, understanding where client data resides, protecting credentials, monitoring vendor incidents, and responding to verified exposure are increasingly important.
By strengthening these practices, MSPs can help clients manage emerging AI security risks while supporting responsible technology adoption and long-term business resilience.