Tool-Use Safety

Tool-use safety refers to the controls that keep an AI system from misusing the software tools it can operate, such as sending emails, editing files, running code, making purchases, or calling external services. As AI agents gain the ability to take real actions on a company's systems and accounts, a mistake or manipulated instruction can produce consequences that are hard to reverse, including data leaks and unauthorized transactions. For compliance and risk teams, this area covers permission limits, human approval steps for high-impact actions, and logging of what the AI did and why, all of which support accountability under emerging AI rules.

1 item

Corporate PolicyUS2026-11-09

Microsoft Agent Security Guide Makes Model Supply Chain a Named Control Requirement

Microsoft has published Secure autonomous agentic AI systems, a guidance document covering how enterprises should select, vet, and monitor the AI models that power their agents. The document addresses model supply-chain governance, adversarial testing, input and output filtering, and controls against unsafe tool use. It directly implicates model inventory programs, version-change approval processes, and ongoing behavioral monitoring in production.

Microsoft agentic AI model supply chain red teaming prompt injection tool-use safety agent security