The company says its latest cybersecurity AI delivers stronger threat detection while reducing costs compared with rival models.
Redmond, Washington: Microsoft has unveiled MAI-Cyber-1-Flash, a new cybersecurity-focused artificial intelligence model that it says delivers stronger threat detection while significantly reducing operating costs compared with competing AI systems.
The model will power Microsoft’s Multi-Model Agentic Scanning Harness (MDASH) platform, as the company intensifies competition with OpenAI, Google, and Anthropic in the rapidly evolving AI cybersecurity market.
The announcement comes as major technology companies race to build AI systems capable of identifying software vulnerabilities, defending against cyberattacks, and helping security teams respond faster to increasingly sophisticated threats.
Designed for AI-Powered Cybersecurity
Unlike general-purpose AI models, MAI-Cyber-1-Flash has been specifically developed for cybersecurity tasks such as vulnerability detection, threat analysis, and security testing.
Microsoft said the model works alongside larger AI systems, including OpenAI’s GPT-5.4, within the MDASH platform.
Instead of relying entirely on expensive frontier models, MDASH automatically assigns tasks based on complexity.
According to Microsoft, approximately 90% of routine security tasks are handled by MAI-Cyber-1-Flash, while only the most complex 10% are passed to GPT-5.4.
The company says this hybrid approach delivers industry-leading performance while cutting operating costs by around 50% compared with using OpenAI models alone.
High Scores in CyberGym Testing
Microsoft said the combined system achieved a 96% score on CyberGym, a benchmark designed to evaluate how effectively AI agents identify, analyse, and reproduce real-world software vulnerabilities.
According to the company, the result outperformed Anthropic’s Mythos model by 12 percentage points.
CyberGym has become one of the key benchmarks for comparing AI cybersecurity systems, measuring their ability to detect bugs, analyse exploits, and simulate realistic attack scenarios.
Microsoft Introduces Project Perception
Alongside the new AI model, Microsoft also announced Project Perception, an autonomous security framework designed to continuously assess and strengthen an organisation’s cyber defences.
The platform operates through three specialised AI agents that mirror how human security teams work.
- Red Team agents simulate cyberattacks by attempting to identify vulnerabilities.
- Blue Team agents analyse the discovered weaknesses to determine their severity and potential business impact.
- Green Team agents recommend or implement defensive measures to fix vulnerabilities and improve security.
Microsoft said the three-agent system creates a continuous feedback loop that constantly discovers, evaluates, and improves an organisation’s security posture.
Project Perception will enter public preview on 3 August.
AI Cybersecurity Arms Race Intensifies
Microsoft’s latest announcement highlights the growing competition among major AI companies to dominate cybersecurity.
As generative AI becomes more capable, security researchers have warned that the same technology helping defenders can also make cybercriminals more efficient by automating vulnerability discovery, phishing campaigns, malware development, and exploit creation.
That has prompted technology companies to invest heavily in defensive AI tools.
Earlier this year:
- Anthropic introduced Project Glasswing and its cybersecurity-focused Claude Mythos model.
- OpenAI launched the Daybreak Cyber Partner Program, followed by GPT-5.6, which it describes as its most advanced cybersecurity model to date.
- Microsoft introduced the MDASH platform in May before expanding it with MAI-Cyber-1-Flash and Project Perception.
Lower Costs Could Benefit Security Teams
One of Microsoft’s biggest selling points is affordability.
The company said the high cost of running advanced AI models has become a major challenge for cybersecurity teams because security operations require constant monitoring and analysis of enormous volumes of data.
By assigning routine work to a smaller, specialised AI model while reserving more expensive models for complex investigations, Microsoft believes organisations can improve security without dramatically increasing computing costs.
As cyber threats continue to evolve and AI-powered attacks become more sophisticated, specialised security models such as MAI-Cyber-1-Flash are expected to play an increasingly important role in helping organisations detect vulnerabilities faster, automate routine security operations, and respond to threats before they can be exploited.
Source: Inc
Read more news and follow us on Instagram
Photo: Adobe Stock
