Microsoft has expanded its bug bounty incentives for security researchers targeting vulnerabilities in Dynamics 365 and Power Platform services, with rewards ranging from $1,250 to $60,000 for qualifying submissions.
The initiative targets security flaws with direct and demonstrable impact in Microsoft-supported cloud products. Microsoft is especially prioritizing cross-tenant vulnerabilities, which are among the most serious risks affecting multi-tenant enterprise platforms.
A cross-tenant flaw could allow an attacker operating from one organization’s Microsoft environment to access data, resources, administrative functions, or business processes associated with another customer tenant.
Microsoft Offers Up to $60,000
Such vulnerabilities can undermine the isolation controls that separate organizations using the same cloud platform.
Microsoft has categorized critical cross-tenant vulnerabilities as high-impact scenarios that qualify for a 100% reward multiplier. Important cross-tenant findings are eligible for a 50% multiplier.
This approach encourages researchers to focus on flaws that could affect multiple organizations or enable access beyond an attacker-controlled tenant.
The expanded scope includes online Dynamics 365 applications, including Sales, Customer Service, Finance, Supply Chain Management, Business Central, Commerce, and Customer Insights.
It also covers Power Apps, Power Automate, Power Pages, Power Admin, AI Builder, Dataverse, and Microsoft Copilot Studio. Selected on-premises Dynamics products are also included in the program.
Microsoft is offering additional bonuses for specific high-impact technical scenarios. A vulnerability that enables privilege escalation into Dataverse through another Power Platform entry point can receive a 20% award multiplier.
Researchers who discover a Dataverse Plugin Sandbox “guest-to-host” escape may also qualify for a 20% multiplier. Such an escape could let code running in a restricted plugin environment cross a security boundary and gain access to a more privileged execution context.
The program also includes AI-focused bounty categories. Critical AI inference manipulation and inferential information-disclosure vulnerabilities can earn rewards of up to $30,000 when they demonstrate substantial security impact.
Microsoft distinguishes these issues from generic AI behavior, making clear that a report must show a real security consequence beyond attacker-controlled content or interactions.
Prompt injection findings without broader security impact are not eligible. The company also excludes reports involving model hallucinations, attempts to obtain system prompts, and content-related AI issues that do not create a meaningful security vulnerability.
Among traditional vulnerability classes, unsafe deserialization of untrusted data and code-injection bugs can receive rewards of up to $20,000.
Critical, high-impact findings involving authentication failures, information disclosure, SQL injection, command injection, server-side request forgery, and improper access control can qualify for payouts of up to $12,000.
Researchers must test the latest fully patched version of an eligible product or service. To receive an award, submissions must be rated Critical or Important and include reproducible evidence establishing both the vulnerability and its impact.
Reports must be submitted through the MSRC Researcher Portal and should include the affected Power Platform or Dynamics environment ID, the testing account username, detailed reproduction steps, proof-of-concept evidence, and the applicable high-impact scenario.
Microsoft excludes vulnerabilities affecting unpatched versions, customer-controlled misconfigurations, user-created applications or content, denial-of-service attacks, cookie replay, subdomain takeovers, and blind XSS.
Researchers are advised to use test tenants, follow Microsoft’s Rules of Engagement, and avoid affecting customer data, privacy, or service availability.
Detect 58% more threats with fresh intelligence from 16K+ organizations. Integrate TI Feeds in you SOC
