Company AI Risk Explorer
Loading the data…
Welcome to the Company AI Risk Explorer

This tool maps what large global companies say in public about AI risk. We reviewed public documents from roughly 200 of the world's largest companies. We coded every risk they acknowledge against the AI Risk Repository taxonomy.

Filter by company, sector, AI role, or risk. Compare what companies acknowledge against expert ratings of severity, sector vulnerability, and role responsibility.

← Back to the overview Company AI Risk Explorer

Companies × risks

Microsoft

Microsoft provides AI infrastructure through Azure (GPU compute, managed model hosting via Azure OpenAI/Foundry Models) and deploys third-party frontier models, especially OpenAI, across customer-facing Copilot products. It develops its own foundation models (Phi family) and specialised models (BioGPT), and influences AI governance through the Frontier Model Forum and congressional testimony.

We reviewed 13 public documents from Microsoft and found it acknowledges 14 of 24 AI risks, including AI security vulnerabilities and toxic content.

Sector: Information

Roles: AI infrastructure provider; AI deployer; General-purpose AI developer; Specialized AI developer

AI risks Microsoft acknowledges

  • AI security vulnerabilities: “protect against indirect prompt injection attacks; ... an attack mechanism where threat actors embed hidden malicious instructions in a grounding data source to circumvent safety guardrails” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)
  • Toxic content: “The Copilot experience can potentially produce other types of inappropriate or offensive content... AI-generated images that potentially contain harmful artifacts such as hate symbols... and sexually charged content that evades sexual-related content filters.” (Transparency Note for Microsoft Copilot)
  • Capability & robustness: “many existing widely used benchmarks are not robust to distribution shifts and other measures of validity over time and fail to translate well to real-world settings” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)
  • False information: “identify several key risks, including the possibility of generating impressions that don't align with findings included in the report-also known as ungrounded content.” (Responsible AI Transparency Report | Microsoft)
  • Discrimination: “Identify and prioritize demographic groups, including marginalized groups, that may be at risk of being differentially impacted by the system... [and] designed to do so in a manner that minimizes disparities in outcomes for identified demographic groups” (Microsoft Responsible AI Standard v2 General Requirements)
  • Disinformation & influence: “Systems that could restrict, infringe upon, or undermine the ability to realize an individual's human rights, e.g., computer vision-based systems deployed in public spaces that could restrict rights to assembly.” (Responsible AI Transparency Report | Microsoft)
  • Overreliance & unsafe use: “Be aware of the risk of overreliance: Overreliance on AI happens when users accept incorrect or incomplete AI outputs... For the end-user, overreliance could result in decreased productivity, loss of trust, product abandonment, financial loss, psychological harm, physical harm, among others. (for example, a doctor accepts an incorrect AI output).” (Transparency Note for Microsoft 365 Copilot)
  • Devaluation of human creativity: “Microsoft has also committed to indemnify and defend customers ... against claims of copyright infringement as a result of the output content generated by these tools ... [to] limit the risk of infringing outputs” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)
  • Weapons & cyberattacks: “tracked high-risk capabilities," which include: Chemical, biological, radiological, and nuclear (CBRN) weapons. A model's ability to provide significant capability uplift to an actor seeking to develop and deploy a chemical, biological, radiological, or nuclear weapon.” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)
  • Unequal performance: “Identify and document any justifiable factors... that account for: 1) any inability to meet any target minimum performance level for any identified demographic group, and 2) any remaining performance differences between identified demographic groups.” (Microsoft Responsible AI Standard v2 General Requirements)
  • Fraud & scams: “...prevent malicious actors from creating and disseminating abusive AI-generated content, including deepfakes, intended to defraud, abuse or manipulate people.” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)
  • Dangerous capabilities: “tracking the emergence of new and advanced AI model capabilities that could be misused to threaten national security or pose at-scale public safety risks” (Responsible AI Transparency Report | Microsoft)
  • Loss of privacy: “The team also used red teaming to test the model's ability to infer sensitive attributes from users' voices. They found that while the model inferred sensitive attributes in 27% of test prompts” (Responsible AI Transparency Report | Microsoft)
  • Governance failure: “As the science of AI evaluations is still developing, there are numerous limitations to current approaches to evaluations... there are gaps in how to systematize, operationalize, and measure relevant theoretical constructs needed to perform reliable testing.” (Microsoft: G7 Hiroshima AI Process (HAIP) Transparency Report)

Documents we reviewed