The common thread in the latest cloud security incidents is not a novel vulnerability but a familiar one: valid credentials. In each case logged recently, attackers gained access through trusted identities and legitimate cloud services, then used that access to reach other customers' data and, in one case, to run up fraudulent charges. The perimeter is no longer a wall to break; it is a set of keys to steal, and the keys are increasingly held by third parties and automated systems rather than by the victim alone.
Access, Not Exploitation
BleepingComputer reported that a threat actor used compromised identity and access management credentials for a Qubole big data platform's AWS account to access and exfiltrate data from multiple customers' AWS environments. That is a cross-tenant compromise achieved through a single set of credentials, not through a flaw in AWS itself. The distinction matters. When an attacker holds valid IAM credentials, the cloud provider's monitoring sees authorized API calls, and the victim's own controls see a familiar identity. The usual signals that flag brute force or exploit attempts do not fire, and the blast radius is determined by what that identity was permitted to do, which in a platform serving many customers is a great deal.
The Third-Party Multiplier
The Railway incident, also reported by BleepingComputer, follows the same logic at a different layer. Railway said hackers exploited vulnerabilities in its infrastructure to steal customer authentication secrets and generate approximately $740,000 in fraudulent AI usage, potentially exposing credentials for up to 100,000 workspaces. Here the initial weakness was in the provider's infrastructure, but the theft was of secrets that grant access to customer resources. The downstream harm landed on customers in the form of fraudulent AI consumption charges, not on the provider alone. Both cases describe a supply chain of trust in which a platform holds the keys to many downstream accounts, and a compromise at the platform propagates outward.
Automation Is Now a Target
Commvault's announcements at its SHIFT 2026 event, as covered by SiliconANGLE, point to where the defensive industry is heading. The company unveiled Recovery Studio, a service that automates the rebuilding of entire business applications after an outage or cyberattack, alongside an AgentRecover update covering recovery for artificial intelligence agents and joint work with Databricks, Nvidia and Zscaler. The recovery focus is telling. Vendors are building for the assumption that destructive access will succeed and that recovery must be fast and automated. But AgentRecover also illustrates a new exposure: AI agents hold credentials and act on their own, and they must be recovered and re-secured like any other identity. The same automation that makes recovery faster makes an agent a valuable target.



