Mate Security introduces Gamebooks, an approach that lets AI agents independently investigate security incidents without unlimited freedom. Organizations define the boundaries in advance, while the agent determines how it conducts the investigation within those parameters.
According to SiliconANGLE, the security startup is thus attempting to strike a balance between traditional automation and fully autonomous AI agents. Traditional security playbooks typically prescribe step-by-step instructions for what to do. This makes them predictable, but also vulnerable to changes in the IT environment. New tools, configurations, or infrastructure can require redesigned workflows.
AI agents are more flexible, but they actually pose risks when allowed to take actions independently. For example, during an investigation, an agent might decide to block an account or shut down a system. Human approval prevents unwanted actions but also slows the automation process.
The goal is key
Gamebooks must combine both models. Instead of defining every investigation step in advance, a Gamebook describes what an agent must investigate and the conditions the investigation must meet. This includes the evidence that must be collected and situations in which the agent must change course. It also specifies which actions are permitted independently and when human approval is required.
This gives the agent the freedom to determine its own investigation path, but it does not operate without limits. An orchestrator selects the relevant Gamebooks based on an incoming alert. Reusable security functions are decoupled from specific vendors’ products.
A separate execution layer handles communication with security tools and other systems. As a result, the AI agents do not need permanent access to these systems. Mate uses its Security Context Graph to gather information about the environment. This graph contains both the current situation and previous investigations and decisions.
The latter is also intended to make the approach less dependent on individual security analysts. When someone leaves, previous decisions and the reasoning behind them remain available. According to Mate, replacing a security product also does not require rebuilding the entire investigation process from scratch. Only the layer that communicates with the relevant technology needs to be adapted.
Investigations improve future detections
Organizations can customize the included Gamebooks or create new procedures in natural language themselves. Existing playbooks can also serve as a starting point. Behind the scenes, Mate handles the technical setup, evaluation, and testing of the AI agents.
Completed investigations are then reused as context. Evidence, correlations, and outcomes are fed into the Security Context Graph. Recurring patterns can trigger new detections, while rules that primarily generate false positives can be refined.
Mate views the recent events surrounding Hugging Face as an example of why speed is crucial in this context. As AI systems become capable of acting autonomously at an ever-faster pace, Mate believes that a security model requiring human approval for every critical step could become a limitation.
Gamebooks is now available on the Mate Security platform. The startup, founded in 2025 by former employees of Wiz and Microsoft, raised $35 million in July in a Series A funding round led by Canaan Partners. This brings its total funding to over $50 million.