Cyberattack on Fireworks AI Exposes Notion Customer Chat Transcripts

A cyberattack targeting the AI platform Fireworks AI has resulted in the exposure of customer data, including billing information, datasets, and chat transcripts belonging to users of online document editor Notion [1]. The incident represents the second data breach experienced by Fireworks AI within the span of two months [1].

Fireworks AI, a platform utilized by developers to access and operate open artificial intelligence models, confirmed that an unauthorized actor breached its cloud provider [1]. According to the company's security advisory, the attacker gained entry to a development system, thereby accessing customer data [1]. Fireworks AI serves tens of thousands of global customers, including AI company Cursor and legal tech firm Harvey [1].

Impact on Notion and Users

Among the affected clients is Notion, which reported that the incident impacted approximately 1% of its customer base [1]. After being alerted by Fireworks AI regarding a security issue, Notion severed all connectivity between its systems and Fireworks on November 3, and subsequently deleted stored data [1].

Notion later confirmed that the intruders stole chat transcripts from Notion AI customers who utilized the Insights feature [1]. These compromised transcripts encompass any model inputs or outputs recorded within the chat history [1]. However, Notion noted that a significant portion of the data remains unattributable, making it impossible to trace the exact source customer for those specific records [1].

Motives and Security Concerns

Fireworks AI stated its belief that the cybercriminals were financially motivated [1]. According to the platform, the hackers aimed to sell tokens obtained by breaching access to large language models, or LLMs [1]. Such illicit schemes allow unauthorized third parties to exhaust AI model usage resources without expending their own assets [1].

The event underscores vulnerabilities associated with technology supply chains, illustrating how reliance on external tech providers can cause security breaches to cascade into downstream customer ecosystems [1]. The latest incident follows an earlier security occurrence at Fireworks AI in October [1].

Fireworks AI is a platform used by developers to access and work with open AI models, serving tens of thousands of customers worldwide, including prominent firms like Cursor and Harvey [1]. Third-party cloud and infrastructure dependencies remain a critical security vector for modern software and AI deployment pipelines [1].

As Indian enterprises and global firms increasingly integrate third-party AI platforms and open models into their workflows, security incidents at infrastructure providers highlight the risks of supply chain vulnerabilities and data spillovers.

The incident may prompt organizations utilizing third-party AI platforms to re-evaluate their API and infrastructure dependencies, encryption practices, and data retention policies when integrating external large language model services.

Open questions remain regarding the full scale of data accessed across all tens of thousands of Fireworks AI customers, as well as the exact methods used by attackers to compromise the cloud provider and development system.