What does NIST’s 2026 agent initiative change?
NIST announced its AI Agent Standards Initiative on 17 February 2026, focusing on confident adoption and interoperable agent systems. Its broader AI Risk Management Framework remains a voluntary risk-management resource. Neither should be described as a federal certificate authorising a particular voice-agent deployment. (nist.gov)
Our recommendation is to translate standards language into testable procurement questions:
- Does the agent have an identifiable system identity?
- Are permissions restricted to the approved workflow?
- Can access be revoked promptly?
- Can the buyer reconstruct which instruction, tool call and version produced an action?
- Can the workflow be stopped without waiting for the supplier’s next release?
What does ISO/IEC 42001 establish—and not establish?
ISO/IEC 42001:2023 concerns AI management systems. It is useful governance evidence, but its subject matter is not a guarantee that a specific appointment booking, opt-out implementation or disclosure satisfies every applicable law. (iso.org)
Ask for the certificate’s scope, covered entity, services and audit status where a supplier claims certification. Then separately test the exact workflow being purchased.
Will US buyers have to replace offshore call centres?
The verified FCC document is a March 2026 notice of proposed rulemaking, not a universal requirement to bring all customer service onshore. It seeks comment on measures including overseas call-volume limits, disclosures and transfers to US-based representatives, with particular attention to communications providers within FCC authority. (docs.fcc.gov)
For BPO buyers, the recommended contract response is a change-control mechanism: identify hosting and staffing locations, allocate responsibility for new requirements, and avoid an unsupported warranty that a US-hosted AI system automatically counts as domestic service under any future rule.