A litigation headline needs an owner and a calculation
Meta’s trillion-dollar exposure figure is a party’s account of maximum-penalty arithmetic. Keep it separate from a requested remedy, a likely outcome and a court award.
The $1.4 trillion figure circulating around Meta's Oakland trial is not a court award. It is Meta's account of maximum-penalty arithmetic.
Berkeley Law's August 17 analysis identifies that distinction and reports a substantially smaller likely request discussed by the states. The court's June 29 pretrial order required the attorneys general to expand their remedies tables so the judge could understand the calculation for each claim.
That order offers a better habit for an AI-governance presentation than the headline does: name the author of the number, identify what is being counted and preserve its procedural status.
Exposure is not an expected loss
A maximum exposure scenario, a party's requested remedy, a settlement and a final award are different objects. A presentation that labels them all “liability” can make an impressive comparison while combining numbers that answer different questions.
I would not infer Meta's communications strategy from the size of its estimate. We do not need to know why a party advances a number to know that the attribution matters. A litigant's calculation should remain a litigant's calculation when it enters a risk discussion.
As an illustration, suppose I am reviewing a hypothetical AI product risk register. Someone adds a large penalty from another company's litigation and applies a rough percentage to our revenue. I would ask them to pause before that result becomes a budget or a board-level forecast.
Which conduct is alleged? Which rule applies? What is the unit of violation? Does the amount describe a theoretical ceiling, a demand or an adjudicated result? Counsel would need to assess whether any of that mechanism applies to our product and jurisdiction.
Without those answers, the figure is context for attention, not a defensible estimate of our loss. Removing it from the model does not mean dismissing the underlying harm. It means finding evidence that can support the decision we are actually making.
Keep product consequences in view
The financial headline can also distract from remedies that change operations. The Berkeley analysis discusses the distinction between monetary penalties and injunctions. A risk review should ask what product behaviour may need to change, without assuming that a particular outcome has already been ordered in this case.
In the hypothetical AI product review, I would therefore maintain a separate account of the decisions the system makes and the people affected by them. That provides a concrete basis for discussing controls, evidence retention and changes to the workflow, even while litigation outcomes remain uncertain.
The court order is useful on another point: it says the court will decide the issues, describing the absence of a binding jury trial. A headline about a jury should not silently become a claim that a binding verdict has settled the dispute.
Steal a small discipline for the next governance deck. Beside every legal-risk number, put the source, speaker or party, date, calculation basis and status. If a field is unknown, keep it visible. Ask the legal team to review the interpretation rather than asking a large number to create urgency on its own.
A careful label will rarely make the slide more dramatic. It will make the resulting decision easier to defend when the next stage of the case produces a different number.
A litigation number enters a risk model with its author, basis and status—not just its magnitude.
