[Fireworks, the AI inference giant with $1 billion in annualized revenue, has been hacked twice, prompting major clients to consider leaving the company] Beating AI News reports that Fireworks, an AI inference and model training platform, has suffered two consecutive hacking attacks. On November 9th, the company acknowledged that some clients' billing information, datasets, and incomplete model weights had been accessed without authorization. Two clients interviewed by Forbes stated that they have already migrated most of their operations away from Fireworks. The first intrusion occurred in early October. Hackers stole Fireworks' internal credentials and obtained API keys used by some clients for model evaluation. At the time, the company stated that it had not yet discovered any access to client models or datasets. On October 24th, hackers again gained access to Fireworks' R&D machines through a compromised third-party cloud service provider, running unauthorized programs and further impacting client data. One client interviewed by Forbes complained that Fireworks is experiencing increasingly frequent hacking attacks, and communication with clients has become very poor. Some of Fireworks' major clients, who pay millions of dollars annually, are considering switching vendors. Fireworks' annualized revenue reached $1 billion in July, with a valuation of $17.5 billion. The company had planned to launch a new round of funding soon, targeting a valuation of over $30 billion. The client attrition could put pressure on this funding. Fireworks stated that the attackers may be linked to a black market that steals and resells large model access limits. Currently, no impact has been found on core services such as online inference, and payment card numbers have not been leaked. The company has suspended some data upload and download functions and has invited Unit 42, a subsidiary of cybersecurity firm Palo Alto Networks, to assist in the investigation. [Original Link]