AIM
Subscribe
Article Complete
AIM
Microsoft Unveils MAI-Cyber-1-Flash, Claims 50% Lower AI Security Costs
Microsoft Unveils MAI-Cyber-1-Flash, Claims 50% Lower AI Security Costs
AI News

Microsoft Unveils MAI-Cyber-1-Flash, Claims 50% Lower AI Security Costs

The launch marks Microsoft’s first in-house AI model built specifically for cybersecurity and software vulnerability detection.

Siddharth Jindal
Published Updated

Key takeaways

  • Microsoft launched MAI-Cyber-1-Flash, its first in-house AI model dedicated to cybersecurity and vulnerability detection.
  • The new model allows MDASH to achieve significant cost reductions while outperforming competitors in cybersecurity benchmarks.
  • MAI-Cyber-1-Flash is designed to handle most vulnerability detection tasks, freeing up larger models for complex issues.
  • The introduction of Perception enhances MDASH's capabilities by using AI agents for continuous monitoring and remediation.
  • Microsoft's model performance is attributed to its design, extensive security data, and the orchestration capabilities of MDASH.
5 min
FollowPreferred SourceGoogle NewsWhatsAppTelegram

Microsoft has introduced MAI-Cyber-1-Flash, its first AI model built for cybersecurity, alongside updates to MDASH, its multi-agent vulnerability identification and remediation platform. 

The company said the new model enables MDASH to deliver “world-class performance at 50% of the cost” of its previous offering while outperforming rival AI systems on a cybersecurity benchmark.

According to Microsoft, the combination of MAI-Cyber-1-Flash and GPT-5.4 achieved a score of 95.95% on CyberGym, a benchmark that evaluates how AI systems identify vulnerabilities across large codebases. The company said this was about 12 percentage points higher than Anthropic’s Mythos model.

Microsoft said MAI-Cyber-1-Flash was built to handle up to 90% of vulnerability detection tasks, allowing MDASH to reserve larger models, including GPT-5.4, for more complex cases. The company said this multi-model approach cuts costs by 50% compared with its earlier MDASH configuration, which relied on GPT-5.4, GPT-5.4 mini and GPT-5.3 Codex.

The company also announced Perception, a new agentic security system within MDASH. Microsoft said Perception uses teams of AI agents to monitor, patch and remediate vulnerabilities continuously, and will expand to additional security workflows using MAI-Cyber-1-Flash.

Microsoft attributed the model’s performance to three factors: the model itself, security data collected over years of operating its platforms, and MDASH's orchestration layer. 

The company said the compact model was derived from its MAI-Thinking-1 family and trained on security-focused data. It also said its security operations generate more than 100 trillion security signals daily across identity, endpoint, cloud and network environments, while MDASH coordinates more than 100 specialised AI agents for vulnerability detection and remediation.

“Three things matter today: Model. Data. Harness,” the company said. “We have jointly optimised our world-class models, our unmatched historic data, and our expert-tuned harness to ensure that our customers have a uniquely powerful security offering.”

Microsoft said MAI-Cyber-1-Flash was evaluated by its AI Red Team, tested through automated and expert-led adversarial exercises, and independently assessed by a third party before release. Through MDASH, customers will also have access to role-based access controls, tenant isolation, encryption, audit logs and sandboxed execution environments without internet access.

The company said its long-term goal is to improve cyber models through reinforcement learning based on real-world security operations. 


Related stories