Definition
The continuous practice of discovering and risk-scoring every AI model, agent, dataset and tool connection an organisation runs, and then fixing the gaps — the AI analogue of cloud security posture management. It surfaces overprivileged identities, misconfigured gateways and shadow AI that would otherwise stay invisible.
Why it matters
Most enterprises do not know how many AI agents and model endpoints they actually run or which ones carry too much access; AI-SPM turns that unknown attack surface into a visible, measurable one.