Authorize
Confirm objectives, owners, scope, access, and stop conditions in writing.
Human-led security service
AI Red Teaming
Challenge agreed AI-system boundaries and safeguards with reproducible misuse scenarios.
What we evaluate
Scope the decision before testing the control.
Challenge agreed AI-system boundaries and safeguards with reproducible misuse scenarios. Scope, access, evidence sources, and safety conditions are agreed before work begins.
Evaluate
Gather and test evidence using practices appropriate to the agreed environment.
Validate
Review observations, uncertainty, business context, and practical remediation priorities.
Brief
Deliver decision-ready findings, owners, dependencies, and recommended next actions.
What you receive
Evidence your team can act on.
System boundary
Documented for decision-making, ownership, and follow-through.
threat scenarios
Documented for decision-making, ownership, and follow-through.
evidence
Documented for decision-making, ownership, and follow-through.
severity rationale
Documented for decision-making, ownership, and follow-through.
mitigations
Documented for decision-making, ownership, and follow-through.
residual limitations
Documented for decision-making, ownership, and follow-through.
Standards-informed approach
Recognized practices, applied to the engagement.
References guide the method and evidence boundary. They do not make this engagement an accredited audit, certification, legal opinion, or formal PCI validation.
NIST AI RMF and GenAI Profile
Used to structure applicable scope, evidence, and recommendations—not to assert certification or compliance.
NIST SP 800-218A
Used to structure applicable scope, evidence, and recommendations—not to assert certification or compliance.
OWASP AISVS and LLM Top 10
Used to structure applicable scope, evidence, and recommendations—not to assert certification or compliance.
ISO/IEC 42001 and 23894
Used to structure applicable scope, evidence, and recommendations—not to assert certification or compliance.
Customer-authorized by design
Start with the outcome and scope.
Our services are human-led, customer-authorized engagements. The Spacetime platform remains purely defensive.
Scope an AI red team →