U.S. and U.K. AI governance reveal the limits of birthright access

At a glance

What this is: This is a C1.ai analysis of how U.S. and U.K. AI governance differ, showing that cultural and regulatory differences make birthright access and static entitlement models increasingly hard to defend.

Why it matters: It matters to IAM, IGA, PAM, and NHI teams because regional AI adoption and contractor governance now shape access design, recertification, and just-in-time controls.

👉 Read C1.ai's analysis of U.S. and U.K. AI governance and access control

Context

The governance problem here is not AI capability alone. It is the assumption that access can be assigned once, left in place, and then managed later through periodic review, even when AI use, contractor involvement, and regional compliance expectations are changing at different speeds.

In the U.K. and Europe, the article describes stronger expectations around data protection, human judgment, and scrutiny of AI policies and training data. In the U.S., deployment is moving faster and oversight is lighter, which creates a mismatch for global identity programmes that try to apply one entitlement model everywhere.

Key questions

Q: How should security teams remove birthright access from AI-adjacent roles?

A: Start by identifying roles that can reach AI tools, contractor data, or regulated records, then remove automatic inheritance of those permissions. Replace them with explicit, task-scoped grants that have an owner, a justification, and an expiry condition. The goal is to make access deliberate and reviewable instead of assumed.

Q: Why do standing contractor entitlements create more risk in multinational AI programmes?

A: Because they outlive the business case that justified them. When contractor access remains active across regions, it can ignore local expectations around data handling, human review, and customer scrutiny. That creates a larger blast radius than the same entitlement would create in a single-jurisdiction model.

Q: What are the signs that AI governance controls are not keeping pace with adoption?

A: Common warning signs include unclear ownership for AI use cases, inconsistent approval processes, limited visibility into where sensitive data enters models, and weak evidence for audits or assessments. Teams also struggle when privacy, security, and legal review happen late or manually, because that usually means governance is reactive rather than embedded in the AI delivery process.

Q: When should organisations prioritise just-in-time access over standing AWS permissions?

A: Prioritise just-in-time access when a role is needed only occasionally, when elevated privileges are high risk, or when engineers need admin rights only for a defined window such as an on-call rotation. It is also the better choice for incident response and outage work because access can be granted quickly and then revoked, reducing the time exposed credentials remain usable.

Technical breakdown

Why birthright access breaks under regional AI governance

Birthright access is the idea that users, contractors, or systems inherit permissions by default and keep them until someone removes them. That model works only when access patterns are stable and governance is centrally enforced. The article shows that AI use complicates this because approval expectations, data handling rules, and operational tolerance for risk differ across the U.S. and U.K., so a single standing-access model becomes misaligned with real-world practice.

Practical implication: replace inherited access with task-scoped entitlement design for AI-adjacent roles and contractors.

Just-in-time access versus standing contractor entitlements

Just-in-time access grants permissions only when a specific task needs them, then removes them when the task ends. Standing contractor entitlements do the opposite: they create long-lived exposure windows that are difficult to justify in environments where access can cross customer data, HR content, and AI-assisted workflows. The article’s contractor examples show why that matters. Once third-party access becomes an entry path, static assignment becomes a governance liability rather than a convenience.

Practical implication: route contractor access through approval-bound, time-limited workflows and track who can reissue access.

Human judgment requirements change the access model

The article notes that European AI governance includes a right to human judgment in decisions affecting people, alongside stricter expectations for transparency around AI-generated content and personal data use. That means access is no longer only about authentication and permission scope. It also touches who may act, who must review, and which outputs require oversight. For identity teams, this pushes governance beyond pure provisioning into decision-path control.

Practical implication: map access policies to the decision process, not just the account, when AI touches employee or customer data.

Threat narrative

Attacker objective: The objective is to exploit long-standing third-party access as a low-friction route into sensitive environments and data.

1. Entry occurs through third-party contractor access, where standing entitlements create a usable path into an organisation.
2. Escalation follows when that access is broad enough to reach sensitive systems or data without just-in-time constraint.
3. Impact lands in the form of prolonged exposure, weak accountability, and a larger blast radius when AI-related workflows or customer data are in scope.

Breaches seen in the wild

• Scania insurance portal breach 2025: An attacker used an external user login, likely stolen by infostealer malware, to take insurance claim documents from a Scania portal.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 150+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Birthright access is the wrong inheritance model for AI governance. The article shows that U.S. and U.K. organisations are operating under different cultural and regulatory assumptions, yet many identity programmes still inherit access by default. That model fails when AI use changes faster than entitlement review cycles. Practitioner conclusion: access should be issued for the task, not assumed from role ancestry.

Regional governance differences are now an identity design problem, not just a policy problem. The U.S. tolerance for faster deployment and the U.K. and Europe’s stronger scrutiny of human judgment and data handling create inconsistent access expectations across the same enterprise. That inconsistency exposes a hidden governance gap: one global entitlement model cannot safely serve every jurisdiction. Practitioner conclusion: identity architecture must carry regional policy logic, not just central provisioning logic.

Contractor access has become the most visible failure mode in multinational AI adoption. The article’s examples point to third-party access as the path where birthright permissions become operationally dangerous. This is not simply overprovisioning. It is the persistence of access beyond the accountability boundary that justified it. Practitioner conclusion: contractor entitlements need lifecycle ownership, not informal extension.

Just-in-time access is moving from optimisation to governance necessity. The article links AI adoption, customer expectations, and human-judgment requirements to a smaller tolerance for standing privilege. That means the old distinction between convenience and control is collapsing. Practitioner conclusion: IGA and PAM teams should treat standing access as the exception, not the default.

Human judgment requirements force identity teams to govern decisions, not only identities. The EU-style expectation that people can challenge or review AI-influenced actions changes what access control is supposed to protect. The control plane must account for who can act, who must review, and when AI output becomes part of a regulated decision. Practitioner conclusion: identity governance must extend into decision accountability.

What this signals

Birthright permissions no longer scale cleanly across AI governance regimes. The U.S. and U.K. examples show that access design now has to absorb regional differences in oversight, data handling, and acceptable human judgment. For practitioners, that means the same role can no longer carry the same standing rights in every market.

Contractor lifecycle governance is becoming the practical control point. Once third-party access can open the door to AI-adjacent systems or sensitive records, lifecycle ownership matters more than whether the original approval was sound. The operational question is no longer who asked for access, but who is responsible for ending it.

Access policy must follow decision policy. If an organisation requires human judgment for certain AI-influenced actions, the identity programme has to know who may act, who must review, and where the final accountability sits. Otherwise the access model and the governance model drift apart.

For practitioners

• Replace birthright permissions for AI-adjacent roles Identify roles that touch AI tools, contractor data, or regulated personal information, and remove automatic inheritance of access. Require task-specific grants with explicit business justification and an owner who can defend the entitlement later.
• Move contractor access into time-bound workflows Give third-party users access through approval-based, time-limited workflows rather than persistent accounts. Revalidate access when the work scope changes, the region changes, or the vendor relationship ends.
• Map regional policy differences into identity rules Separate access logic for jurisdictions that treat human judgment, data sovereignty, and AI output disclosure differently. Do not rely on one global entitlement template when local regulatory expectations diverge.
• Tie AI governance to decision ownership Record who approved the use of AI, who reviewed the output, and who is accountable for the final action when personal data or employee-related decisions are involved. That prevents access from becoming detached from responsibility.

Key takeaways

• The article frames AI governance as an access problem as much as a policy problem, especially where standing entitlements outlast the context that justified them.
• Regional differences in regulation and security culture make one-size-fits-all entitlement models increasingly unreliable for multinational organisations.
• Just-in-time access, contractor lifecycle control, and decision accountability are the controls most likely to reduce the exposure described in the post.

Key terms

• Birthright Access: The baseline set of entitlements that a user should receive by default because of role, department, or another stable attribute. It is a governance construct, not a blanket permission model. The control challenge is proving that the baseline stays current as jobs, applications, and ownership change.
• Just-in-Time Access Request: Just-in-Time Access Request is a pattern that grants access only when it is needed and only for the duration required. It reduces standing privilege by making access temporary, policy driven, and task scoped. This approach is especially useful for contractors, sensitive systems, and short-lived operational work.
• Contractor Entitlement: A contractor entitlement is an access grant issued to a third-party worker or vendor account. These entitlements are often the weakest link in identity governance because they can persist beyond the work period, cross regional policy boundaries, and create accountability gaps if not tied to a clear owner.
• Human Judgment: Human judgment is the decision-making layer that evaluates whether AI output is accurate, safe, and appropriate for use. It matters because AI can accelerate creation, but people still need to interpret results, challenge assumptions, and decide what should move forward.

What's in the full article

C1.ai's full blog post covers the operational detail this post intentionally leaves for the source:

• The first-hand conversation with Abraham Ingersoll and Alex Bovee on U.S. versus U.K. AI governance
• The specific examples of contractor access and retailer disruption discussed in the article
• The article's discussion of human judgment, data sovereignty, and AI policy expectations across regions
• The cultural observations behind differing attitudes to AI adoption and risk tolerance

👉 C1.ai's full post covers the cultural differences, contractor access examples, and AI policy tensions in more detail.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.