Microsoft Project Perception targets AI-powered cyber threats

Microsoft said it has launched Microsoft Project Perception, an agentic security system built to counter cyberattacks that use artificial intelligence and automated agents. The company said in a Monday blog post that the system converts security signals into real-time protections, reflecting the growing demand for tools that can respond at machine speed.

Microsoft described the product as a way to use AI against AI-enabled threats. In the company’s account, effective defense depends on a continuing view of how attackers identify openings, how defenders judge risk and how protections can be updated as conditions change.

What is Microsoft Project Perception?

Project Perception is a security system that coordinates specialized AI agents, models, signals and response tools to assess cyber risk and strengthen defenses over time, according to Microsoft. The system is designed as a closed loop: it detects possible weaknesses, evaluates whether they matter and applies improvements to reduce exposure.

Microsoft said the system uses three types of agents. Red team agents look for vulnerabilities before attackers exploit them. Blue team agents investigate findings and decide which issues amount to material risk. Green team agents apply fixes and harden security controls.

The company said those agents operate together rather than as separate tools. The aim is to create an automated process in which discovery, assessment and remediation feed into one another, giving organizations a system that learns from new signals and adjusts protections.

How the system works

Project Perception combines multiple layers, Microsoft said: signals and sensors, security context, AI models, an orchestration layer and actuators that translate decisions into protective action. The orchestration layer, which Microsoft called a harness, coordinates the work of the agents and models.

The system also uses what Microsoft described as a multi-model architecture. That includes frontier models and cyber-focused specialized models, with the company saying it will select models for individual security tasks while considering quality and cost.

Microsoft said it also plans to keep developing its own specialized models for security work as part of that approach. The company framed Project Perception as part of a broader effort to improve outcomes as attackers adopt AI tools.

Why Microsoft is expanding AI security products

PYMNTS reported July 15 that Microsoft’s cybersecurity business was building more AI security products, scaling back some traditional security offerings and consolidating engineering teams. According to that report, Microsoft was responding to customer demand for defenses against AI-enabled attacks and competing for spending also flowing to AI companies including Anthropic and OpenAI.

The commercial case for AI in security also extends to fraud prevention. A PYMNTS Intelligence report, “Where Payments Decisions Happen: How Issuer Data Is Powering the Next Era of Commerce,” found that 42% of issuers said AI helped them avoid more than $5 million in fraud losses in recent years.

For Microsoft customers, Project Perception adds another AI-centered offering to the company’s security portfolio. For the broader market, it signals how large technology providers are repositioning cyber defenses around continuous assessment, automated response and specialized models as AI changes the cost and speed of attacks.

This story draws on original reporting from PYMNTS.