Leaked AWS keys enable attackers to hijack Amazon Bedrock models and drain victim accounts
Security firm FortiGuard Labs reported that stolen long-lived IAM credentials were used to authorize unauthorized generation of text, creating massive billable inference costs on compromised cloud accounts.
Threat actors are exploiting leaked Amazon Web Services credentials to hijack generative AI models on Amazon Bedrock. On September 3, security company FortiGuard Labs analyzed an AWS account compromise involving a long-lived Identity and Access Management access key carrying AdministratorAccess permissions. The stolen credential allowed the attacker to create a new IAM user, subscribe to foundation models through AWS Marketplace, and invoke those models. This process generated billable inference activity on the compromised account without valid authorization. Security experts describe this technique as LLMjacking, where attackers abuse valid cloud identities to consume expensive model services rather than steal data. The operator used the fresh identity to accept marketplace agreements required for foundation models before making calls to generate text. These calls produced direct charges against the victim's AWS account. The attack chain began with a single administrator-level access key that did not expire automatically, allowing unauthorized subscription and usage.
