Quick answerWhen a single live interaction spans two jurisdictions whose rules genuinely conflict, such as one market requiring a disclosure the other prohibits, the agent needs a pre-built conflict resolution hierarchy that names which jurisdiction's rule controls in each specific scenario, rather than defaulting to whichever rule the agent encounters first or trying to satisfy both simultaneously. Build this hierarchy with legal counsel before it is needed, since resolving it live, mid-conversation, is not a decision an agent should make on its own.
This is not the outbound-contact question or the localization question
Cross-border consent law for proactive AI agent contact governs whether the company can reach out to a customer first. Localizing an AI agent for a market with different consumer protection law governs how the agent operates once it is launched in a single market. Both assume the interaction lives cleanly inside one jurisdiction's rules. A genuinely cross-jurisdiction interaction, a customer physically located in one country, an account domiciled in another, a transaction governed by a third contract's choice-of-law clause, does not fit that assumption, and the agent can end up straddling two rule sets that actively conflict inside the same conversation.
What a real conflict looks like in practice
A common shape: one jurisdiction requires the agent to proactively disclose a specific piece of information as a consumer protection matter, while a second jurisdiction's privacy rule restricts sharing that same category of information without a separate, explicit consent step the conversation has not yet reached. Satisfying one rule risks violating the other in the same message. This is meaningfully different from the data-location question covered in data residency and sovereignty when your model provider is abroad, which is about where data sits, not about which live-interaction rule takes precedence when two rules point in opposite directions.
Building a conflict hierarchy in advance, not live
The agent cannot be the one that resolves a genuine legal conflict in the moment, and it should not try. What it can do is detect the conflict pattern (two specific jurisdiction flags present on the same interaction, matched against a known conflict table built with counsel) and follow a pre-decided resolution, such as defaulting to the more restrictive rule, routing to a human before continuing, or applying a jurisdiction-specific script variant built for exactly this overlap. The hierarchy has to be built and signed off before the conversation happens; asking the agent to weigh two regulatory regimes against each other live is asking it to practice law without a license, badly.
Flagging the gap in the hierarchy itself
No conflict table will anticipate every combination of markets a company operates in, especially as the business expands into new regions. The agent needs an explicit fallback for the case where it detects a cross-jurisdiction interaction that does not match any pre-built resolution: pause, route to a human with jurisdiction expertise, and log the specific combination so legal can add it to the hierarchy going forward. Treating every unmatched case as low-risk and letting the agent proceed on its own judgment defeats the entire point of building the hierarchy in the first place.
FAQ
How common is this in practice versus a theoretical edge case? It comes up most often with mobile or relocated customers, subsidiaries operating under a different jurisdiction than the parent contract, and cross-border B2B accounts where the buyer and the billing entity sit in different countries.
Should the agent just default to the stricter jurisdiction's rule whenever it detects any conflict? Defaulting to the stricter rule is a reasonable fallback for genuinely unmapped conflicts, but for known, common conflicts it is worth building an explicit resolution rather than always taking the most restrictive path, since that can create its own compliance gaps in the other direction.
Who owns maintaining the conflict hierarchy over time? Legal or compliance should own the content of the hierarchy, with the AI platform team owning the detection logic that routes conversations into it, similar to how ownership splits in most AI agent governance processes.