Follow us on Google News
Artificial intelligence infrastructure startup Fireworks AI has suffered its second cyberattack within a span of weeks, resulting in unauthorized access to files containing customer records. The Nvidia-backed company disclosed the latest security incident to its clients on Monday following a breach that occurred on October 24. An unauthorized third party infiltrated systems operated by an external cloud hosting vendor where the firm’s computing machines were stationed. Early disclosures indicate that an intruder managed to gain unauthorized access to internal systems and stored files. According to a person familiar with the development, the breach exposed tens of thousands of corporate files, a portion of which contained sensitive customer details.
Breach Targets External Cloud Provider Hosting Core Machines
The startup confirmed via a company blog post that the perpetrator specifically targeted its third-party cloud service provider. During the unauthorized intrusion, the attacker was able to access several hosting environments as well as specific internal company files. While a substantial volume of documentation was exposed during the October 24 incident, the firm has not publicly revealed the precise number of clients whose records were compromised, nor has it defined the exact nature of the accessed data.
In response to the compromise, a spokesperson for Fireworks AI stated that the enterprise is actively coordinating remediation measures to assist affected organizations. The company confirmed that it is in active communication with key institutional clients, including productivity software maker Notion and developer platform Cursor, to manage the immediate impact of the security compromise.
Major Enterprise Clients Rely on Core Infrastructure Services
Fireworks AI provides foundational infrastructure services that allow enterprises to deploy, run, and scale artificial intelligence models across commercial operations. According to corporate figures on the startup’s official portal, more than 10,000 corporate clients rely on its technological ecosystem. High-profile organizations using its platform include legal technology provider Harvey, workspace application developer Notion, and code generation platform Cursor.
Because the company provides backend architecture that hosts live software models, the security of its storage volumes and internal technical configurations remains critical to enterprise operations. Breaches involving third-party cloud hosting vendors present extensive cross-platform risks, as unauthorized entry into an upstream provider’s systems can jeopardize data belonging to downstream institutional clients. However, current records do not verify whether individual end users of Notion, Cursor, or Harvey had their personal details extracted during the intrusion.
Repeated Network Breaches Reveal Vulnerabilities Across AI Supply Chains
The recurrence of a second successful network penetration in recent weeks has intensified scrutiny over Fireworks AI’s internal controls and its reliance on outside hosting facilities. Two consecutive security breaches within a short timeframe present serious data privacy risks and heighten operational concerns across the company’s enterprise user base.
The incident underscores the broader cybersecurity challenges confronting AI infrastructure providers that utilize multi-tier supply chains. Artificial intelligence ventures frequently rely on external technical platforms, specialized data centers, and third-party hosting partners to support intensive computing requirements. When a single external vendor is compromised, the exposure can ripple through connected platforms, endangering numerous dependent businesses simultaneously. Fireworks AI stated that it is currently collaborating with impacted clients to evaluate the full scope of the breach and implement preventative security remedies. Detailed technical assessments regarding the cause of the second breach, the precise inventory of affected files, and the potential misuse of data remain pending.
