Wiz counts at least nine AI developer platforms hit by security incidents since October

Cybersecurity company Wiz counts at least nine AI developer platforms that disclosed security incidents during October, among them Fireworks AI, Railway, Lightning AI, OpenCode and Warp, and has been adding further cases in November; the latest, Tensorlake, disclosed on November 8, according to the update to its report of November 9.

According to Wiz, all public reports from the affected companies point to targeted exploitation aimed at stored environment variables, especially those granting access to AI services, although no single vulnerability or dependency links all the cases. Fireworks, OpenCode and Warp have reported indications of an “agent swarm” attack, according to the same report.

Wiz says it has observed downstream abuse of enterprise keys stolen in one of those breaches. In one of the two cases it describes, about 800,000 calls to AI models on Amazon Bedrock were made with a single key over thirty hours; in the other, about 1.7 million calls in thirteen hours, spread across a dozen Anthropic Claude models, according to the company.

Zeabur, one of the affected platforms, will shut down as a result of the incident, according to Wiz. The report also cites, as accounts posted on social media, the alleged theft of about 371,000 dollars in cryptocurrency using keys stored in Railway.

Developing story: the information comes from a single source, the Wiz report, and the affected companies' own statements could not be consulted. According to Wiz itself, most of the incidents are still being analysed with no public technical details, and no public body has confirmed in the pages consulted that the attacks were carried out by AI agents.