A Security Expands Autonomous Security Testing to Internal Networks and Private Cloud
A Security has expanded its autonomous offensive security and remediation platform to include internal networks and private cloud environments with the launch of Continuous Internal Infrastructure Testing.
The new capability extends the company's existing external infrastructure and application testing by enabling organizations to assess how an AI-powered adversary could exploit their internal infrastructure after gaining initial access. The platform measures the potential business impact of a compromise, identifies attack chains and autonomously remediates weaknesses while continuously validating that fixes remain effective.
According to A Security, organizations that test external exposures and internal networks separately can overlook how multiple vulnerabilities and misconfigurations may combine to create a path to full network control. Threat actors with valid credentials and knowledge of internal infrastructure can chain together weaknesses to bypass security measures such as endpoint detection and response (EDR), multifactor authentication (MFA) and network segmentation.
Continuous Internal Infrastructure Testing is intended to help security teams identify and address these attack paths, providing visibility into how attackers move from the perimeter into internal networks.
“Weaponized AI lets attackers move through enterprise networks faster than ever by adapting to defenses and linking together weaknesses that are typically addressed in isolation,” said Omer Gull, chief product officer at A Security. “We already help Fortune 1000 organizations find and close the external exposures attackers use to get in, and with Continuous Internal Infrastructure Testing, we now cover everything from the perimeter to the crown jewels within a single platform.”
Gull added that customers wanted to understand what an attacker could access from a compromised account, address the weaknesses that make those paths possible and continuously verify that the vulnerabilities remain closed as their environments change.
The expansion comes as threat actors increasingly use autonomous AI agents to identify weaknesses, adapt their tactics and pursue alternative routes to critical systems. A Security pointed to recent research involving the China-nexus group VAPOR PANDA, which reportedly used AI agents in a cyber espionage campaign targeting more than 200 organizations across at least 14 countries.
The agents conducted reconnaissance, harvested credentials and moved laterally through targeted environments while changing attack paths when an approach failed. A Security noted that, unlike scripted tools that stop when an initial attempt is blocked, autonomous agents can continually adjust their tactics until they identify a viable route into an environment.
The company said the availability of capable open-weight models that can be downloaded without verification or content filters makes these capabilities accessible to a broader range of attackers.
The new offering includes several capabilities intended to help organizations identify, remediate and validate internal attack paths.
Testing begins with a single domain user account, such as one obtained through phishing or stolen credentials. A Security's agents map the network and private cloud environment before pursuing objectives established by the customer. The agents adapt their approach during testing and evaluate newly disclosed vulnerabilities as they become available.
The platform also identifies attack chains by connecting vulnerabilities and misconfigurations, including issues without a Common Vulnerabilities and Exposures (CVE) identifier. These evidence-backed chains are mapped to MITRE ATT&CK to show how multiple low-severity weaknesses can combine to create a path to a full takeover. A Security then autonomously remediates the identified chains and continues retesting them to verify that they remain closed.
Internal testing operates continuously alongside the company's external infrastructure and application testing, providing security teams with a consolidated view of how attackers could gain access and move through internal networks and private cloud infrastructure.
Lightweight connectors support remote and segmented sites while customers maintain control over testing scope and approvals. Layered Guardrails are also included to prevent destructive actions.
For regulated and air-gapped environments, organizations can deploy A Security within their own AWS, Azure or Google Cloud accounts and direct AI calls exclusively to approved models. Findings, evidence and credentials remain within the boundaries organizations already govern, and the deployment does not require internet access.
Continuous Internal Infrastructure Testing is currently available through an Early Access Program. Organizations interested in participating can visit A Security's Early Access Program page for more information and to apply.
