Microsoft Launches Global AI Red Team Alliance Across Six Continents

Microsoft is expanding AI security testing beyond its internal walls with a new global initiative that acknowledges a critical gap: no single organization possesses the multilingual expertise, regional knowledge, and domain specialization needed to identify all high-risk failure modes in frontier AI systems.

The company's AI Red Team announced the External Red Team Alliance (EXTRA), a formalized program providing unrestricted funding to 18 university labs across six continents while building a distributed network of external specialists who can participate directly in security assessments of advanced AI systems.

According to details first reported by Microsoft Security, the initiative represents a fundamental shift in how the company approaches AI safety testing as models become more capable and attack surfaces expand beyond traditional concerns like prompt injection.

Why it matters

AI red teaming increasingly requires expertise that spans cybersecurity operations, multilingual harms, alignment failures, and domain-specific abuse patterns that vary by geography and language. Just as modern cybersecurity depends on coordinated vulnerability research and independent security researchers who identify risks vendors miss, advancing AI safety requires contributions from experts across institutions, disciplines, and regions. Microsoft's acknowledgment that internal teams cannot fully replicate this breadth of expertise signals a maturation in how the industry approaches frontier model security.

Two-part structure addresses research and operations

EXTRA operates through two distinct components. The first provides unrestricted gifts to university labs already investigating unresolved questions in AI safety, intentionally avoiding predetermined deliverables or product requirements. The funding aims to strengthen independent research capacity globally rather than direct outcomes toward Microsoft's needs.

The second component focuses on operational collaboration, building a network of specialists who can participate directly in red teaming highly specialized areas. This includes researchers and practitioners with expertise in specific attack classes, languages, cultural contexts, or technical domains that internal teams may not adequately cover.

Research areas funded through the program span both offensive and defensive dimensions: examining how AI systems can be attacked, manipulated, or abused in operational environments, as well as exploring how AI systems can assist defenders and improve cyber operations.

Expanding beyond corporate security boundaries

Microsoft's AI Red Team observed that meaningful testing of advanced systems requires broader participation from researchers and practitioners operating outside traditional corporate security boundaries. The initiative reflects the company's Frontier Governance Framework approach, which combines rigorous internal governance with external support and collaboration.

"Managing frontier AI risk requires more than internal safeguards," said Natasha Crampton, Chief Responsible AI Officer at Microsoft. "It requires continuous engagement with experts who understand how these systems behave across different technical, linguistic, and cultural contexts."

Mike Yeh, VP and Deputy General Counsel for Customer Security and Trust at Microsoft, emphasized the national security implications: "Understanding how these systems can be misused, and identifying risks before they become real-world threats, requires expertise that spans institutions, disciplines, and borders."

The program's unrestricted funding structure is deliberate. Rather than directing research toward specific outcomes, Microsoft aims to create stronger long-term collaboration between academia and operational AI security teams while strengthening the broader ecosystem needed for trustworthy AI development.

The External Red Team Alliance details were announced by Microsoft Security on the company's security blog.