Rights and contracts
Can licensing pricing data to AI create antitrust risk?
By SourceX Editorial · Reviewed by Noah Loul ·
Short answer
Licensing pricing data to an AI developer can create antitrust risk when current, nonpublic prices could reach competitors or feed a shared pricing tool. For 3PLs, freight companies and distributors, the safer path is to exclude current rates and margins, license only aged or aggregated price history, and bar the buyer from using the records to recommend prices to others.
Key takeaways
- Antitrust concern centers on competitors exchanging current, nonpublic prices, not on licensing operational records as such.
- Rate cards, spot quotes, bid responses, carrier pay and margin reports are the most sensitive records in a logistics archive.
- There is no fixed safe harbor for how old or how aggregated shared pricing must be, so counsel judges each package on its facts.
- Exception notes, routing decisions and service records can usually be licensed once dollar amounts are masked.
- The license should bar the buyer from using your records in pricing tools offered to companies in your market.
Why can pricing data raise antitrust questions?#
Pricing data raises antitrust questions because US antitrust law, including the Sherman Act and the FTC Act, treats exchanges of competitively sensitive information between competitors as a possible route to coordinated prices. The concern is not that a company shares data at all. The concern is that rivals end up knowing, or acting on, each other's current and future prices.
Government actions and private lawsuits over algorithmic pricing software have followed a recognizable pattern. Competitors feed nonpublic, current pricing into a common tool, and the tool recommends prices back to each of them. Plaintiffs argue that the software does the coordinating that the competitors could not lawfully do in a room together, even if the competitors never speak to each other.
Licensing records to an AI developer is a different activity. The risk appears when your prices could travel through the buyer to a competitor, for example because the buyer builds a freight pricing product, a quoting assistant or a market benchmark that other carriers, brokers or distributors use.
Which logistics records carry pricing risk?#
The logistics records that carry pricing risk are the ones that reveal what you charge, what you pay and how you bid. A typical archive spread across a TMS, a WMS and an ERP such as NetSuite or Epicor mixes these with operational records that carry little pricing signal once the amounts are removed.
The overlap is often smaller than general counsel expect. What AI developers tend to value in logistics records is operational reasoning: why a load was rerouted, how a short shipment was resolved, what a dispatcher did after a missed appointment. Those records can usually be separated from the price fields that sit beside them.
| Record type | Where it usually lives | Pricing sensitivity | Common handling |
|---|---|---|---|
| Customer rate cards and contract rates | TMS rate tables, ERP price lists, spreadsheets | High: current and customer-specific | Exclude, or license only after counsel approves aging and aggregation |
| Spot quotes and bid responses | TMS quote history, RFP files, email | High: shows bidding strategy | Exclude recent periods; mask amounts in older records |
| Carrier pay and settlements | McLeod or other TMS settlement records, accounts payable | High: reveals cost structure | Usually exclude |
| Margin and profitability reports | ERP reporting, BI dashboards | High | Usually exclude |
| Accessorial and surcharge charges | Freight bills and invoices | Medium | Keep the event description, remove the amount |
| Exceptions, claims and service notes | TMS, WMS and ticketing tools | Low once amounts are masked | Usually licensable after privacy preparation |
| Routing, dock and inventory decisions | WMS, routing tools, telematics such as Samsara | Low | Usually licensable |
What safeguards reduce the risk?#
The main safeguards are aging, aggregation and excluding current pricing, backed by contract terms that control how the buyer uses the records. They work best in combination, because each one closes a different path by which a price could reach a competitor.
There is no fixed safe harbor for how old or how aggregated shared pricing must be. The federal antitrust agencies have withdrawn their older guidance that treated certain historical, pooled exchanges as presumptively safe, and they now review the market, the data and the likely effect case by case. A cutoff that suits a fast-moving spot freight market may be wrong for long-term distribution contracts.
- Exclude current pricing: leave out active rate cards, open bids, renewals under negotiation and any forward-looking pricing plans.
- Age the history: license price-bearing records only if they are older than a cutoff that counsel sets for your market, not a rule of thumb.
- Aggregate where a signal must stay: replace individual rates with ranges or indexes that blend many lanes, customers and periods.
- Mask amounts in operational records: swap dollar fields in exceptions, claims and invoices for placeholders while keeping the event text.
- Restrict use: bar the buyer from using the records to build, tune or benchmark pricing recommendations for others in your market.
- Limit access: keep the records with the buyer's model development team, with no resale and no transfer to a pricing product group.
Does it matter who the AI buyer is?#
The identity and business of the AI buyer matter a great deal, because the same records carry different risk in different hands. A developer training a general-purpose model on records from many industries is a different setting from a company building pricing or quoting software for freight, warehousing or distribution.
Ask the buyer directly what the model will do, who will use its outputs and whether the buyer or an affiliate sells pricing, benchmarking or marketplace services in your sector. Record the answers in the license as representations, so a later change of plan becomes a contract question rather than a surprise.
Models can sometimes reproduce passages from their training data. Masking amounts before delivery matters for that reason too: a contract promise not to expose prices is weaker than records that never contained them.
Illustrative: a freight brokerage separates rates from operations#
Illustrative: a fictional freight brokerage with about 150 employees runs its loads through a TMS, tracks check calls in the same system and handles shipper and carrier email in shared inboxes. Its archive holds customer contract rates, spot quotes, carrier pay, load tenders, check-call notes and years of exception threads about missed pickups, detention and damaged freight.
A model developer wants the records to train a dispatch assistant. During diligence, the brokerage learns that an affiliate of the developer sells a freight rate benchmarking product to brokers and shippers. Outside counsel flags contract rates, spot quotes, carrier pay and margin reports as the material that could raise antitrust questions in that setting.
The brokerage licenses check-call notes, exception threads and load events with every rate, accessorial amount and pay figure replaced by a placeholder, and excludes rate tables, quotes and settlements entirely. The license names dispatch assistance as the only permitted use, bars any use in pricing or benchmarking products, and keeps the records away from the affiliate.
What should the license say about pricing data?#
The license should define excluded pricing data, limit how the buyer may use what remains and give you a way to confirm compliance. Clear definitions matter more than broad promises, because a vague exclusion invites a dispute about what counted as pricing.
| Clause | What it does | Why it matters for antitrust |
|---|---|---|
| Excluded data | Lists rate cards, quotes, bids, carrier pay, margins and forward pricing as outside the license | Keeps the most sensitive fields out of scope by definition |
| Permitted use | Names the model purposes the records may serve | Prevents a quiet shift into pricing products |
| Competitor restriction | Bars use in tools that set or recommend prices for others in your market | Closes the shared-algorithm path |
| Output controls | Requires measures against reproducing figures from the records | Reduces the chance that a model surfaces your numbers |
| No onward transfer | Bars resale, sublicensing and transfer to affiliates without consent | Keeps the records away from pricing teams |
| Attestation and audit | Lets you request confirmation of use and deletion | Gives you evidence if questions arise later |
How SourceX handles pricing fields#
SourceX treats pricing as a rights and preparation question inside the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. During Rights, rate cards, quotes, carrier pay and margin fields are flagged for the supplier's counsel. During Preparation, amounts the supplier chooses to exclude are removed or masked before anything is delivered.
The SourceX Evidence Packet records which fields were excluded, the permitted use and the supplier's release authorization, so the supplier's counsel and the buyer work from the same record. SourceX does not give antitrust advice; the supplier and its counsel decide what pricing material, if any, is licensed.
Frequently asked questions
Is it illegal to license old freight rates to an AI company?
No general rule makes it illegal, and none makes it automatically safe. The answer depends on how old and specific the rates are, how concentrated your market is, who the buyer is and what the model will do. Antitrust counsel assesses those facts deal by deal, which is one reason many suppliers simply leave price fields out.
Do publicly posted prices carry the same risk?
Publicly posted prices, such as published tariffs or list prices on a website, generally carry less risk than nonpublic contract rates, because competitors can already see them. The risk returns when public figures travel together with nonpublic discounts, customer names or volumes, so check what else sits in the same records.
Does masking prices make the records less useful to AI developers?
Usually less than suppliers fear. Developers training models on logistics work tend to want the sequence of events and decisions: the exception, the options considered, the action taken and the result. Dollar amounts add little to that sequence, and placeholders can show that a charge existed without revealing what it was.
Should outside antitrust counsel review the package?
If price-bearing records are in scope at all, a review by counsel with antitrust experience is sensible, especially in concentrated regional markets or where competitors use the same pricing software. If prices are fully excluded and masked before delivery, a general rights review may be enough, but that is counsel's call.
Can the buyer use our records to build a freight rate model?
Only if the license allows it, and suppliers should be cautious about allowing it. A rate model trained on your pricing and then sold to brokers, shippers or carriers in your lanes comes close to the shared-algorithm pattern that enforcers and plaintiffs have challenged. Name that use as excluded in the permitted-use clause.
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.