Notion AI Data Breach Exposes Customer Chat Transcripts After Fireworks AI Cyberattack
Share
A cyberattack on AI infrastructure provider Fireworks AI has exposed private AI chat transcripts belonging to Notion customers. Notion confirmed that fewer than 1% of its workspaces were affected after attackers accessed data held by its third-party AI provider.
The exposed information included customer conversations processed through Notion’s Insights feature, which helps workspace administrators understand how employees use AI. Although most of the affected transcripts couldn’t be linked to specific users or workspaces, attackers could still access their contents.
The incident raises concerns about the security of third-party AI services and the risks businesses face when they share information with external technology providers. It also highlights a gap between contractual data protection commitments and the security of the infrastructure used to process customer information.
What information did the attackers access
The compromised data consisted of AI chat transcripts associated with Notion’s Insights feature. The feature helps administrators understand how employees interact with AI tools across their workspaces.
Notion clarified that most of the affected transcripts didn’t contain identifying information that could connect a conversation to a specific user or workspace. The exposed dataset also didn’t include tool call results. However, the attackers could access the actual conversation content.
This creates a potential privacy risk for businesses and their employees. AI conversations can contain internal discussions, customer information, business plans, technical details, and other confidential material. Even when transcripts don’t include names or account identifiers, their contents may reveal sensitive information.
Notion hasn’t disclosed the exact number of affected customers. The company has confirmed that the incident affected fewer than 1% of its workspaces.
How the Fireworks AI breach happened
The incident involved two separate security events at Fireworks AI during October 2026. The first incident occurred in early October, when attackers gained access to internal credentials and compromised environment variables containing API keys and authentication tokens associated with a limited number of customer accounts.
A second attack took place later that month. Attackers compromised a third-party cloud provider that hosted Fireworks AI’s research and development machines. They used this access to reach Fireworks’ development environment and execute an unauthorized workload.
The second incident exposed customer information, including resource metadata, billing details, datasets, and incomplete model weights belonging to some Fireworks customers. Notion was among the affected customers because it relied on Fireworks to process conversations for certain AI services.
Fireworks contained the second incident on October 24. However, the company didn’t notify Notion about the potential exposure until November 2. Subsequent investigation revealed that the compromised dataset contained customer information.
How Notion responded to the breach
Notion took several steps after receiving information about the incident. Its investigation also revealed that the initial dataset contained more than synthetic testing information, prompting further action to protect customer data.
The company followed this timeline:
- November 2: Fireworks informed Notion that attackers had accessed a dataset. Notion initially identified the material as synthetic testing data.
- November 3: Notion disconnected its systems from Fireworks, deleted data held by the provider, and stopped further communication with its systems.
- November 6: Fireworks informed Notion that customer data might also have been accessed.
- November 8: Fireworks confirmed that the dataset contained customer information.
- November 9: Notion began notifying affected customers.
- November 11: Notion revoked exposed API keys discovered in affected users’ messages during its investigation.
Notion also disabled the Insights feature and stopped using Fireworks to serve open-weight AI models. The company stated that the incident didn’t affect other Notion features.
What should affected Notion users do
Notion has advised affected customers to treat the exposed information as compromised, even when transcripts don’t identify individual users or workspaces. Workspace owners should review the available incident information and take appropriate steps to protect credentials and sensitive business data.
The company recommends the following precautions:
- Review exposed conversations: Workspace owners should use the Security incident review option in Notion’s desktop web interface to access the affected information.
- Rotate compromised credentials: Change passwords, API keys, access tokens, and other credentials that employees may have entered into Notion AI conversations.
- Check sensitive information: Review exposed transcripts for confidential business information, customer records, internal discussions, and other sensitive details.
- Follow Notion’s notifications: Affected workspace owners should follow the instructions provided by Notion and review any additional security updates from the company.
Notion says its investigation is ongoing and it will notify additional customers if it identifies further exposure.
Why the breach raises concerns about third-party AI security
The Fireworks AI incident highlights the security challenges businesses face when they integrate external AI services into their operations. Companies increasingly depend on third-party AI providers for productivity tools, customer support, software development, data analysis, and internal workflows. These integrations can introduce additional security dependencies because information may pass through several systems before reaching its intended destination.
In this incident, Notion had a contractual zero-data-retention commitment from Fireworks AI. However, attackers still accessed a dataset containing customer chat transcripts processed for the Insights feature.
This distinction is important. A zero-data-retention agreement doesn’t automatically protect information from unauthorized access while a service processes it. Businesses must also understand how vendors handle data, secure their infrastructure, manage access credentials, and respond to security incidents.
The incident also highlights the importance of reviewing the security practices of AI infrastructure providers. Organizations should assess third-party vendors before integrating their services and regularly review the permissions and data access granted to them.
Employees also need clear guidelines on what they can share with AI assistants. Passwords, API keys, customer records, financial information, and confidential business documents shouldn’t be entered into AI tools without appropriate security controls.
To Sum Up
The Fireworks AI breach demonstrates how a security incident at one technology provider can affect customers of another company. Although attackers didn’t directly compromise Notion’s infrastructure, they accessed customer conversations processed through an external AI service. For businesses adopting AI tools, the incident reinforces the need to look beyond the security of the applications they use. Organizations must also examine the providers supporting those applications and understand how they protect customer information. As companies expand their use of AI, third-party vendor security, data handling practices, and access controls must remain central to their cybersecurity strategies.
Sources
BleepingComputer: Notion says Fireworks breach exposed customer AI chat transcripts
Notion Official Security Advisory: Fireworks AI Security Incident
