By J.R. Howell

In Uber Technologies, Inc. v. City of Seattle, the Ninth Circuit affirmed the denial of a preliminary injunction against a Seattle ordinance that requires app-based delivery platforms to tell workers what conduct can cost them access to work. One provision of the ordinance captures the stakes of the appeal. A platform may take a customer’s rating into account when evaluating a delivery. What Seattle forbids is a policy under which the number derived from customers’ ratings is, by itself, sufficient reason to deactivate the worker. The customer supplies the rating. The company decides what the rating will cost the worker.[1]

That distinction goes some distance toward answering a question the court had no occasion to decide. When a platform invokes Section 230 in response to a claim that it used control over access to exclude competition, whose decision would a judgment hold it responsible for, and what would the platform have to change to comply? A court that answers those questions can protect a platform’s discretion over third-party speech while still examining a commercial condition the platform imposes on the people who depend on it. A plaintiff who gets that far has a separate burden under antitrust law. The loss of one account establishes very little, standing alone, about the alternatives available to everyone else.

Seattle regulated the grounds on which a worker could lose access

Seattle’s concern was that delivery workers could lose a source of income under standards they had little opportunity to discover until after the platform applied them. The City also questioned what an aggregate score could tell a platform about the worker. A customer who takes a ride observes how the driver drives. A customer who receives a grocery order ordinarily sees the result of the trip. Seattle was entitled to take account of that difference and of its concern that customer bias could enter the score. The ordinance at issue covers the delivery services operated by Uber and Instacart. The rideshare comparison was the court’s explanation for why a delivery rating may tell the company relatively little about safety.[2]

The ordinance requires a written policy sufficiently specific for a worker to know what conduct can lead to deactivation and how to avoid it. The grounds must be “reasonably related” to the company’s “safe and efficient operations.” Seattle gives those terms content. A company cannot adopt a policy that penalizes a worker for rejecting offers, including a specified proportion of offers, or for discussing compensation or working conditions with other workers. The ordinary process calls for an investigation, confirmation of a violation, consistent application of the rule, and a proportionate penalty. A company may act before completing its investigation where the allegations concern egregious misconduct. An allegation that a courier threatened a customer may call for immediate action. A record showing that the same courier declined five orders could be entirely accurate and still supply no lawful ground for deactivation under this provision.[3]

Uber, its subsidiary Portier, and Instacart challenged the requirements governing the content and disclosure of the policy. They did not challenge the separate prohibition against unwarranted deactivations. The district court treated the effect on speech as incidental to the regulation of conduct and denied preliminary relief. The distinction between that reasoning and the Ninth Circuit’s ultimate decision is important. The panel withdrew its March opinion. In the superseding October opinion, Judge Richard Clifton, joined by Judge Susan Graber, assumed that the ordinance regulated speech and examined whether the companies were likely to prevail on that premise.[4]

The disagreement concerned what the company was required to say

For the majority, the identity of the listener helped determine the character of the speech. The company had to tell the people performing deliveries the terms on which they could continue to do so. It could identify the City as the source of a restriction and express its disagreement with the restriction alongside the policy. The worker would still know the rule. The panel distinguished laws requiring social media companies to report to the government on their treatment of controversial public subjects. Seattle required communication to a party to the economic relationship about the conditions of that relationship.[5]

Under Zauderer, as applied in the Ninth Circuit, a requirement to disclose factual and uncontroversial commercial information must bear a reasonable relationship to a substantial governmental interest and must not impose an unjustified or undue burden. Once a company has adopted its policy, the majority reasoned, it can accurately tell a worker which acts will cost that worker access to the app. A single webpage could supply the notice. As for the companies’ argument that Seattle also restricted what the policy could say, the majority reasoned that a rule authorizing an unwarranted deactivation concerned conduct the City had made unlawful. The panel also rejected the vagueness challenge. A company could disagree about the safety value of customer ratings and still understand precisely what the word “solely” prohibited.[6]

Judge Mark Bennett agreed that the law gave fair notice and that Zauderer applied to the compelled commercial disclosure. He located the difficulty one step earlier, when the company has to write the rule. Formulating a policy that reconciles safety with efficiency requires the company to make a judgment. Telling a worker what an existing policy says may be entirely factual even though compelling the company to formulate that policy requires it to decide what safety demands. Seattle separately required a notice of statutory rights and made the policy requirement independently enforceable. In Bennett’s view, the majority also reasoned in a circle insofar as the policy itself helped determine which deactivations were unlawful. He found serious questions on the merits and would have remanded for the district court to reconsider all four preliminary injunction factors.[7]

The force of Bennett’s objection depends on which exercise of judgment the City has required. A company might reasonably prefer to use a ratings cutoff. Once Seattle has forbidden that practice, a statement to a worker that the company cannot deactivate her on that ground tells her something she can use, irrespective of the company’s opinion about the wisdom of the law. I find the majority persuasive on that point. Bennett’s objection carries greater weight where the government leaves the substantive choice to the company and then requires it to formulate and defend its judgment. The difference lies in what the government asks the speaker to do.

Publishing a rating and using it to withdraw work are separate acts

Consider the same customer rating in a Section 230 case. A claim that a service should have removed a customer’s review because it falsely accused a worker of misconduct would require the service to answer for its decision to publish information supplied by the customer. Section 230(c)(1) addresses precisely that relationship between the defendant and the information. Now suppose the worker accepts that the review is accurate and seeks to show that the platform automatically withdrew access to work when an aggregate score crossed a threshold. The company could leave every word and every rating where the customer put them and still change the rule that deprived the worker of access. That difference supplies a reason to examine the company’s eligibility decision separately. It is my application of the statutory text to Seattle’s example. The parties did not litigate Section 230 in Uber.[8]

A plaintiff cannot create that distinction simply by alleging that a publishing decision cost it customers. In Marshall’s Locksmith Service Inc. v. Google, LLC, locksmiths claimed that search companies conspired to display fraudulent competitors as nearby businesses, diverting customers and inducing legitimate businesses to spend more on advertising. The pleaded claims included monopolization and restraint of trade. Yet the false assertion that a locksmith operated at a particular address came from the business that supplied the listing. Converting the address into a map pin did not supply the falsehood. The D.C. Circuit applied Section 230(c)(1) to the claims before it. The plaintiffs forfeited on appeal their argument about alleged manual alteration of results. The opinion therefore supplies no answer to a properly preserved claim based on a platform’s material contribution to the deception.[9]

Where the platform sells the product that competes with the product it filters, a different provision of Section 230 may be at issue. Section 230(c)(2)(A) protects specified good-faith decisions to restrict access to objectionable material. Subsection (B) addresses providing others with the technical means to restrict access. In Enigma Software Group USA, LLC v. Malwarebytes, Inc., a security software company alleged that a direct rival had configured its filtering product to block the plaintiff’s programs in order to divert its customers. Malwarebytes maintained that the programs used deceptive tactics. At the pleading stage, the Ninth Circuit held that the provider of the filtering tool could not obtain dismissal under subsection (B) on the premise that disapproval of a competing product for anticompetitive reasons made the product “otherwise objectionable.” Enigma’s claims arose under state law and the Lanham Act. The decision allowed it to pursue those claims without establishing that the disputed designation was false or that Malwarebytes had monopolized a market.[10]

A security company may know more about a competing product’s defects than anyone outside the business that made it. Judge Johnnie Rawlinson’s dissent in Enigma deserves attention for that reason. The statute assigns the assessment of objectionable material to the provider or user and contains no express disqualification for a provider that competes with the producer of the material. A court should take a substantiated security assessment seriously. In my view, the relevant inquiry is what prompted the designation, what testing supports it, and whether the company applied the same criterion to its own product or to comparable products in which it had no financial interest. Those are questions on which an excluded rival and a provider acting to protect users may both benefit from a record more informative than the label applied to the software.[11]

A condition on access can deprive a rival of the means to compete

Seattle’s protection for workers who decline delivery offers identifies a more precise competition question than the mere fact of deactivation. Suppose a delivery platform conditions continued access to its orders on a courier’s refraining from accepting work from a competing service. The condition would matter to the other service even if the platform never removed a review, changed a rating, or published anything about the courier. If enough couriers depend on the first platform for work, the condition could deprive an entrant of the delivery capacity it needs to serve customers. A court could examine that condition and its effects while leaving the platform free to publish every customer’s assessment. This hypothetical describes an exclusionary mechanism that could be investigated. There was no finding in Uber that either company imposed such a condition.

The Supreme Court confronted that mechanism long before a platform could enforce a rule by disabling an account. The newspaper in Lorain Journal Co. v. United States monitored the customers of a competing radio station and refused advertisements from local merchants who continued to buy time on the station. Merchants who depended on the paper to reach customers had reason to abandon radio advertising even if they would otherwise have purchased both. The resulting loss of advertising revenue threatened the station’s ability to compete. On the evidence of an attempt to monopolize, the Court sustained an injunction against use of that condition. The decree left the newspaper with discretion to reject advertisements on other grounds. A similar distinction would permit a court to address a prohibition on working with a rival while preserving the platform’s ability to remove a courier who presents a genuine danger.[12]

A courier’s lost income would establish the personal stakes of exclusion. A monopolization claim would require evidence of monopoly power in a relevant market and conduct by which the defendant acquired or maintained that power. In the example above, that inquiry would include how many workers are bound by the condition, what proportion of the available delivery capacity they supply, whether rival services can recruit elsewhere, and whether the restriction is necessary to complete an order already accepted. Requiring a courier to finish a delivery undertaken at noon presents a different competitive question from barring that courier from accepting a rival’s work for the rest of the day. A private plaintiff seeking damages would also have to connect its own injury to the aspect of the conduct that makes the restraint harmful to competition.[13]

The written policy at the center of Uber has value to a worker before there is an antitrust case to bring. It gives the worker a standard against which to test the company’s stated reason for withdrawing access. It also makes it possible to ask whose interests a condition serves. If the rule requires a courier to complete the order she accepted, the platform can explain the obligation by reference to the customer waiting for the delivery. If the rule reaches the work she takes after that delivery is complete, the question for counsel is what operational need would justify extending the restriction to a transaction between the courier, another customer, and another service. It is at that point that the terms governing one worker’s account may begin to govern the conditions under which an entire rival business can operate.

---

[1]     Uber Techs., Inc. v. City of Seattle, Nos. 25-228 & 25-231, slip op. at 6–10, 20–24 (9th Cir. Oct. 29, 2026) (affirming the denial of preliminary relief and distinguishing sole reliance on an aggregate rating from consideration of other evidence).

[2]     Uber, slip op. at 8–10, 19–21 (describing the City’s findings and explaining why a delivery customer and a rideshare passenger have different opportunities to observe a driver).

[3]     Seattle, Wash., Ordinance 126878, § 8.40.050.A, .C–.D (Aug. 14, 2023) (setting out the policy requirements, grounds Seattle excludes, investigation procedures, and exceptions in the appendix to the opinion).

[4]     Uber, slip op. at 6–12, 22 (withdrawing the earlier opinion, identifying the provisions challenged, and affirming on the assumption that the ordinance regulated speech).

[5]     Uber, slip op. at 12–17 (distinguishing reports on controversial public issues from a company’s communication of its terms to its workers).

[6]     Uber, slip op. at 15–22 (applying the commercial disclosure standard, addressing the alternative restriction argument, and finding that the ordinance gave fair notice). Zauderer v. Off. of Disciplinary Couns., 471 U.S. 626, 650–53 (1985) (sustaining a requirement to disclose the client’s potential responsibility for costs in advertisements for contingent-fee representation).

[7]     Uber, slip op. at 27–36 (Bennett, J., dissenting in part) (distinguishing disclosure of an existing fact from compelled formulation of a policy and proposing remand for reconsideration of preliminary relief).

[8]    47 U.S.C. § 230(c)(1) (protecting a provider or user of an interactive computer service from being treated as the publisher or speaker of information provided by another information content provider). Uber, slip op. at 19–21, 23–24 (explaining the permissible use of information accompanying a rating and reproducing the prohibition on reliance solely on an aggregate metric).

[9]     Marshall’s Locksmith Serv. Inc. v. Google, LLC, 925 F.3d 1263, 1265–72 (D.C. Cir. 2019) (affirming dismissal of claims based on third-party location information and addressing forfeiture of the manual-manipulation argument).

[10]   47 U.S.C. § 230(b)(2)–(4), (c)(2)(A)–(B) (identifying competition and effective filtering among the statutory policies and defining the protected activities). Enigma Software Grp. USA, LLC v. Malwarebytes, Inc., 946 F.3d 1040, 1045–52, 1053–54 (9th Cir. 2019)(reversing dismissal where a filtering provider allegedly blocked a direct competitor’s software for anticompetitive reasons).

[11]   Enigma, 946 F.3d at 1054–55 (Rawlinson, J., dissenting) (objecting that the text provides no basis for distinguishing filtering decisions by competitors).

[12]   Lorain J. Co. v. United States, 342 U.S. 143, 148–49, 152–56 (1951) (upholding relief directed at a publisher’s use of access to newspaper advertising to prevent merchants from purchasing radio advertising).

[13]   United States v. Microsoft Corp., 253 F.3d 34, 50–51, 58–59 (D.C. Cir. 2001) (en banc) (per curiam) (requiring monopoly power and exclusionary conduct and explaining how competitive effects and procompetitive justifications are assessed). Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 487–89 (1977) (requiring a private damages plaintiff to show injury flowing from the feature of the conduct that makes it anticompetitive).