1 in 6 fleets report AI tools leaked supplier or customer data

How often do fleet AI tools expose confidential pricing data?

One in six supply chain and logistics professionals said their company experienced an incident in the past year in which supplier or customer information was exposed through an artificial intelligence tool, according to a survey of 250 managers and executives fielded in October 2026 by InstaLILY and Pollfish. The survey defined an incident as a situation in which supplier or customer data was exposed, shared, or accessed incorrectly through an AI tool. It did not assess severity, meaning the findings do not distinguish between minor internal mistakes and serious breaches.

The exposure rate comes as fleets and distributors increasingly give AI systems access to sensitive commercial information. Seventy-eight percent of respondents said their company's AI tools could access customer pricing and margins. Sixty-seven percent said the tools had access to supplier pricing and rebate information. The survey targeted companies already using AI in daily operations.

What AI systems have access to in fleet operations

The survey reflects a broader pattern in logistics technology. AI tools now handle tasks that require access to confidential contract terms, including quote generation, margin analysis, and supplier negotiation support. For fleets, that can include carrier contract rates, fuel surcharge formulas, accessorial pricing, and volume rebate tiers. For distributors, it includes customer-specific pricing, supplier cost structures, and rebate agreements.

Sixty-three percent of respondents said their company had been asked by a supplier or customer how it protects their information. For 30%, those inquiries had influenced a decision involving a software provider, vendor, or AI tool. That suggests trading partners are beginning to audit how their data is handled by the AI systems their customers and suppliers deploy.

Confidence vs. reality in AI data controls

Eighty-seven percent of respondents said they were very or somewhat confident their AI tools kept supplier and customer information appropriately separated. At the same time, 63% acknowledged some overlap between supplier and customer data across their company's platforms, including AI tools. The gap between confidence and acknowledged overlap suggests many organizations may not fully understand how their AI systems handle data boundaries.

"Speed means very little if a faster answer costs you a relationship that took years to build," said Adam Isenberg, head of distribution at InstaLILY. "AI working in distribution should know that one customer's contract price never belongs on another customer's quote. Buyers should make vendors prove that before they sign."

The survey did not independently verify the reported incidents. The results reflect respondents' reported knowledge and experiences, not forensic audits of actual data exposures.

Why AI tools leak data in supply chain operations

The most commonly cited obstacle to separating supplier and customer information was that existing AI tools were not designed with that requirement in mind, cited by 22% of respondents. Cost and a lack of technical expertise followed at 21% and 20%, respectively.

That design gap is not unique to logistics. IBM's 2025 Cost of a Data Breach report found that 13% of surveyed organizations had experienced a breach involving AI models or applications. Among those organizations, 97% reported lacking proper AI access controls. The report also found that breaches involving unauthorized AI tools, commonly called shadow AI, added an average of $670,000 to breach costs.

For fleets, the risk is not just financial. A data exposure that reveals one customer's negotiated rate to another customer can destroy pricing leverage and damage long-term relationships. A leak of supplier rebate terms to a competitor can undermine procurement strategy.

What fleets want from AI vendors

Nine in 10 respondents said their employers would be more likely to select an AI tool specifically designed for distribution and supply chain operations if it offered stronger controls over supplier and customer data. Ninety-two percent said data separation was a high or moderate priority for their company over the next year.

The report recommends testing AI systems against routine business scenarios, including whether a tool can prepare a customer quote without exposing confidential supplier pricing or contract terms. It also recommends establishing when human approval is necessary before AI systems act on information.

Fleets evaluating AI tools for freight procurement, quote generation, or margin analysis should ask vendors to demonstrate data separation in live scenarios. A tool that can generate a customer quote should not be able to surface the supplier cost structure that underlies that quote unless the user is authorized to see both. A tool that analyzes carrier contract performance should not be able to expose one customer's accessorial rates to another customer's quote.

What this means for fleet software decisions

The survey results suggest that AI data security is moving from a compliance checkbox to a competitive differentiator. Fleets that can demonstrate strong data controls to their customers and suppliers may gain an edge in contract negotiations. Fleets that cannot may face direct questions from trading partners about how their AI systems handle confidential information.

For fleet managers evaluating AI tools, the takeaway is to test data boundaries before deployment. Ask the vendor to show how the system prevents cross-contamination of customer pricing, supplier rebates, and contract terms. Ask for audit logs that show who accessed what data and when. And establish clear policies for when AI can act autonomously and when it must surface a recommendation for human approval.

The 16% exposure rate in the survey is self-reported and unverified, but it is high enough to warrant scrutiny. A fleet that deploys AI without testing data separation is betting that its vendor got the design right. The survey suggests that bet does not always pay off.