Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:21-cv-00135 — Sharon Johnson Coleman , Judge .

ARGUED JANUARY 28, 2026 — DECIDED JULY 13, 2026

Before HAMILTON, MALDONADO, and TAIBLESON, Circuit Judges .

HAMILTON, Circuit Judge . We review in this appeal objections by two class members to the district court’s approval of a settlement. The plaintiff class asserted claims based on the Page 2use of biometric data taken from public sources on the internet for facial recognition. We find no inherent substantive problems with two challenged features of the settlement, first the absence of injunctive relief, and second monetary relief in the form of what amounts to an equity stake in the defendant. But we cannot get past a key procedural problem in the settlement process.

The settlement provides far greater monetary benefits for members of several favored subclasses based on state residency, leaving the most meager benefits for members of only the nationwide class. Differences in the laws of different states may well make t hose differences in benefits reasonable, but no representative of the disfavored nationwide class endorsed the allocation of monetary benefits. Class representatives who are members of the favored subclasses cannot also represent the disfavored nationwide class concerning the allocation of monetary relief. Class-action settlement requires structural assurances of fair and adequate representation, and those were lacking here. We therefore vacate approval of the settlement and remand for further proceedings.

Defendant Clearview AI, Inc. operates “a search engine for faces.' The company scrapes photographs of individuals from public websites, including social media accounts, and analyzes them using artificial intelligence to produce 'facial vectors' reflecting the geometry of a person’s facial features. Searching this database with a photograph of an unidentified person returns results of all other photographs of that person in the database, linked to the websites on which they were found. Depending on the information available on those web- Page 3sites, a search can therefore reveal a wealth of information about a person, including not only their identity but also, depending on where the photographs were taken, personal relationships and political and religious affiliations, among other facts.

This technology has a wide range of applications. For example, Clearview's customers originally included private companies that wanted to identify people in surveillance camera footage. According to plaintiffs, Clearview also at one time granted database access to its political allies, including certain members of Congress. Since, at the latest, a settlement of a separate lawsuit with the American Civil Liberties Union, however, Clearview’s customers have included only federal and state government agencies and their contractors.

This case arises from several putative class-action suits filed after an exposé on Clearview appeared in The New York Times in January 2020. Plaintiffs filed in the Northern District of Illinois and the Southern District of New York against Clearview, a Delaware corporation headquartered in New York. Other defendants included Clearview's co-founders, defendants Hoan TonThat and Richard Schwartz; defendant Rocky Mountain Data Analytics LLC, a New Mexico company that allegedly served as a shell entity for Clearview; defendant Thomas Mulcaire, Clearview’s general counsel and a vice president of Rocky Mountain; and Macy’s, one of Clearview’s customers.

Pursuant to 28 U.S.C. § 1407, the Judicial Panel on Multidistrict Litigation transferred eleven such cases for coordinated pretrial proceedings in the Northern District of Illinois. After appointment of interim lead class counsel, Page 4Plaintiffs filed a consolidated class -action complaint asserting the following claims:

• · On behalf of a Nationwide Class of all individuals in the United States whose biometric data is or was in Clearview’s database, claims for a declaratory judgment and for unjust enrichment.
• · On behalf of an Illinois Subclass of Illinois residents, several claims under the state's Biometric Information Privacy Act (BIPA).
• · On behalf of a California Subclass of California residents, claims under the state's Unfair Competition Law, statutory commercial misappropriation of identity, common-law right of publicity, and the state constitutional right to privacy.
• · On behalf of a New York Subclass of New York residents, a claim under the state's civil rights code.
• · On behalf of a Virginia Subclass of Virginia residents, claims under statutory commercial misappropriation of identity and the Virginia Computer Crimes Act.

The case was litigated vigorously through motions to dismiss and discovery, producing over five hundred docket entries by the end of 2022. Sometime that year, the parties entered settlement negotiations and retained as mediator the Honorable Wayne Andersen, a retired federal district judge. This first round of negotiations fell apart because Clearview, an undercapitalized startup, was not in the financial Page 5condition to make a large and immediate payment or to provide the other forms of relief the Class demanded.

The parties returned to mediation in early 2023. This time they agreed in principle that any settlement would be based on payment through what amounted to an equity stake in Clearview. By that time, the American Civil Liberties Union had reached a settleme nt with Clearview in a separate case. Under the ACLU settlement, Clearview was permanently enjoined from allowing database access to private parties except in compliance with the Illinois BIPA. Clearview was also required to establish an opt-out program for Illinois residents.

After months of negotiations shepherded by Judge Andersen, the parties agreed to a settlement. The settlement included appointment of a settlement master, the Honorable Sidney Schenkier, a retired federal magistrate judge. Under the settlement, upon an initial public offering or a 'liquidation event'-merger, consolidation, or sale-of Clearview, the Class would receive a payment equivalent to a 23% equity stake in Clearview as of September 6, 2023, 'subject to the same dilution by future investments as for the founders.” Alternatively, the courtappointed settlement master may in lieu of such payment (a) sell the settlement stake to a third party for a “commercially reasonable price” or (b) make a cash demand for 17% of Clearview’s “Generally Accepted Accounting Principles (GAAP) recognized revenue” from the date of final approval of the settlement by the district court until the date of such demand. Such a demand would need to be made no later than September 30, 2027.¹ The agreed relief is only Page 6monetary in nature. Clearview has not committed to any additional limits on its business activities. The settlement stake would be divided among the Class with ten shares to each member of the Illinois Subclass, five shares to each member of the California, New York, and Virginia Subclasses, and one share to each member of only the Nationwide Class.

None of the eight original class representatives agreed to the settlement. Lead Class Counsel strongly implied in the motion for preliminary approval that these refusals were based on strategic disagreements with the representatives’ attorneys, who “variou sly proposed settling for injunctive relief only (allocating the available cash to attorney’s fees), bankrupting Clearview by continuing to litigate … and/or pursuing the defendant class of Clearview customers.” None of the representatives were actually represented by Lead Class Counsel by that point because an attorney had left the firm and took those clients with him. Accordingly, Lead Class Counsel replaced them with four new representatives. At every point, all named class representatives who approved the settlement were members of one of the favored state -specific subclasses.

The district court granted preliminary approval of the settlement and invited objections and exclusions. Sixteen objections were received from members of the Class, and a coalition of state attorneys general filed an amicus brief opposing Page 7the settlement. After a hearing, the district court found the settlement to be fair, reasonable, and adequate and granted final approval on March 20, 2025. See Fed. R. Civ. P. 23(e)(2). Objectors Robert Weissman and Rick Claypool are members of the Nationwide Class and have appealed.

The district court had jurisdiction under the Class Action Fairness Act, 28 U.S.C. § 1332(d). We have jurisdiction under 28 U.S.C. § 1291. We agree with Objectors that the lack of separate class representatives for the Nationwide Class requires that the se ttlement be vacated. For the sake of judicial economy, we also address Objectors’ contentions about the lack of injunctive relief and the uncertain nature of the equity stake settlement, as otherwise those issues might well form the basis of a successive appeal. To be clear, however, this is a general remand, and this opinion does not imply that any future settlement needs to be substantially similar to the one we review here.

We review a district court’s approval of a class-action settlement for an abuse of discretion. In re Southwest Airlines Voucher Litig. , 799 F.3d 701, 711 (7th Cir. 2015). Yet in practice, our review can be quite searching. Pearson v. NBTY, Inc. , 772 F.3d 778, 780 (7th Cir. 2014) (characterizing appellate review of classaction settlements as “far from pro forma”). The district court has considerable discretion, but it must be apparent that the court carried out its role as effectively a fiduciar y to the class. That role demands from a district court “the highest degree of vigilance in scrutinizing proposed settlements of class actions.' Reynolds v. Beneficial National Bank , 288 F.3d 277, 279 (7th Cir. 2002).

Federal Rule of Civil Procedure 23(e) instructs a district court considering a classaction settlement to “approve it only after a hearing and only on finding that it is fair, reasonable, and adequate.” As amended in 2018, Rule 23(e) provides that in reaching its decision, the court must consider whether:

(A) the class representatives and class counsel have adequately represented the class; (B) the proposal was negotiated at arm’s length; (C) the relief provided for the class is adequate, taking into account: (i) the costs, risks, and delay of trial and appeal; (ii) the effectiveness of any proposed method of distributing relief to the class, including the method of processing classmember claims; (iii) the terms of any proposed award of attorney’s fees, including timing of payment; and (iv) any agreement required to be identified under Rule 23(e)(3); and (D) the proposal treats class members equitably relative to each other.

Long before the 2018 amendment, this court had crafted a list of factors to guide the fairness determination: '(1) the strength of the case for plaintiffs on the merits, balanced against the extent of settlement offer; (2) the complexity, Page 9length, and expense of further litigation; (3) the amount of opposition to the settlement; (4) the reaction of members of the class to the settlement; (5) the opinion of competent counsel; and (6) stage of the proceedings and the amount of discovery completed.” Wong v. Accretive Health, Inc. , 773 F.3d 859, 863 (7th Cir. 2014), quoting Gautreaux v. Pierce , 690 F.2d 616, 631 (7th Cir. 1982).

The district court’s analysis tracked our case law, which originated before the 2018 amendments establishing subsections (A) through (D) of Rule 23(e)(2). For future reference, and for the sake of consistency, courts may now, going forward, find it most he lpful to frame their analysis in terms of the factors as set forth in Rule 23(e). As the advisory committee explained the 2018 amendment: 'The sheer number of factors [in circuits’ lists] can distract both the court and the parties from the central concerns that bear on review under Rule 23(e)(2).' Any difference between the standards is not dispositive in this case, however, given the nature of the settlement and the arguments raised on appeal.

Objectors first attack the settlement as not 'fair, reasonable, and adequate' because it would not stop Clearview's challenged conduct and would release claims that they say might have achieved that goal. Objectors appear to identify these claims as BIPA, the Nationwide Class’s unjust enrichment claim, and unpled state constitutional and common law right-toprivacy claims. We conclude that fair settlements of these claims did not require injunctive relief.²

Page 10Before addressing each category of claims, we note a faulty premise underlying these objections. “Because ‘the essence of settlement is compromise,’ courts should not reject a settlement ‘solely because it does not provide a complete victory to the plaintiff s.'' In re AT&T Mobility Wireless Data Services Sales Litig. , 270 F.R.D. 330, 347 (N.D. Ill. 2010) (St. Eve, J.) (alteration omitted), quoting first EEOC v. Hiram Walker & Sons, Inc. , 768 F.2d 884, 889 (7th Cir. 1985), and then Isby v. Bayh , 75 F.3d 1191, 1200 (7th Cir. 1996). Adequacy under Rule 23(e) is quite different from whether a plaintiff who prevails on the merits lacks an “adequate remedy at law” so as to justify injunctive relief. See Romper Room Inc. v. Winmark Corp. , 60 F. Supp. 3d 993, 997 (E.D. Wis. 2014) (no “adequate remedy at law” where damages were “unlikely to provide complete relief”). In any event, the absence of injunctive relief did not prevent the district court from approving this proposed settlement.

First, BIPA could not have served as the basis for nationwide injunctive relief because only the Illinois Subclass asserted BIPA claims. Also, the ACLU settlement already permanently enjoined Clearview from disclosing biometric information to private parties except in compliance with BIPA. That settlement also required Clearview to establish an opt-out program for Illinois residents. See ACLU v. Clearview Page 11AI, Inc. , No. 2020-CH-04353 (Ill. Cir. Ct. May 11, 2022). While that injunction might not have provided complete relief to the Illinois Subclass, the settlement here would have had to offer additional injunctive relief to be anything other than 'superfluous.' See Pearson , 772 F.3d at 785.

Second, the unjust enrichment claim asserted by the Nationwide Class was speculative. Plaintiffs do not allege Clearview obtained its biometric data through hacking, social engineering, 'catfishing,' or any means other than scraping and analyzing photographs already posted and publicly available on the internet. The Objectors have not identified even one class-action biometric data privacy case in which a court has entered judgment on an unjust enrichment claim and awarded injunctive relief for that claim, let alone case law showing that their claims were so strong that they could not fairly be settled without injunctive relief. The remedy for unjust enrichment is ordinarily a money judgment for or disgorgement of the amount of the defendant's unjust enrichm ent. Perhaps a sufficiently 'flexibl[e]' state court could grant something akin to injunctive relief by awarding some clever bundle of “rights of ownership … in specifically identifiable property in the hands of the defendant.' See Restatement (Third) of Restitution and Unjust Enrichment § 49 cmt. a, ch. 7 topic 2 intro. (A.L.I. 2011). But that possibility again does not show that a class settlement without injunctive relief could not be approved.

That possibility raises another problem for the Objectors. The elements of and remedies for unjust enrichment claims differ significantly among the states. Those differences can make certification (in the absence of a settlement) difficult if not improper, and they can make fashioning an appropriate Page 12nationwide injunction just about impossible. See In re Bridgestone/Firestone, Inc. , 288 F.3d 1012, 1015 (7th Cir. 2002) (collecting cases refusing certification of nationwide classes raising warranty, fraud, and products-liability claims because of differences among state laws; “No class action is proper unless all litigants are governe d by the same legal rules.”); Siegel v. Shell Oil Co. , 256 F.R.D. 580, 583–86 (N.D. Ill. 2008) (St. Eve, J.) (denying motion to certify nationwide class bringing unjust enrichment claims due to significant variations in state laws), aff'd on other grounds, 612 F.3d 932 (7th Cir. 2010).³

Third, Objectors point to the release of unpled constitutional and common-law right-to-privacy claims. As Objectors recognize, this argument really gets at the scope of the release, not the lack of injunctive relief, since claims not actually brought could not have served directly as the basis for any relief. Generally, class-ac tion settlements may, and often do, release unpled claims arising from the same facts as the claims actually brought . A settlement may release claims without identifying the released claims in detail and regardless of whether they would have been “presentable in the class action.” See Williams v. General Electric Capital Auto Lease, Inc. , 159 F.3d 266, 273– 74 (7th Cir. 1998) (emphasis omitted), quoting Class Plaintiffs v. City of Seattle , 955 F.2d 1268, 1287 (9th Cir.

Page 131992). After all, most settling defendants expect to buy peace with respect to all claims arising from the relevant events, regardless of how they might be labeled.

The released claims here appear to be just as speculative, and with likely just as much variation among the states, as the unjust enrichment claims. At the risk of understatement, state courts can reach different results in applying state constitutional right-to-privacy provisions to the controversial issues of the day. Compare Planned Parenthood South Atlantic v. State , 438 S.C. 188, 216–17, 882 S.E.2d 770, 785–86 (2023) (plurality opinion), and Planned Parenthood of Montana v. State ex rel. Knudsen , 2025 MT 120, ¶ 127, 422 Mont. 241, 296, 570 P.3d 51, 89 (2025), with Planned Parenthood of Southwest & Central Florida v. State , 384 So. 3d 67, 87–88 (Fla. 2024). Similar variations in tort law are common.

Lead Class Counsel made a strategic choice to pursue a settlement based on an equity stake alone, without injunctive relief. Clearview was not likely to agree to a settlement that surrendered its business model, and the ACLU settlement blunted the Class’s strongest claims for injunctive relief. So, obtaining an injunction here almost certainly would have required going to trial, which the Class had no guarantee of winning and which may well have exhausted Clearview's assets before it could pay anything at all, according to Judge Andersen. The district court did not abuse its discretion by approving a settlement without injunctive relief.

Objectors also challenge the settlement based on three related arguments against the fairness, reasonableness, and adequacy of the monetary relief it offers.

Page 14First, Objectors are correct that there is no guarantee that Clearview will ever have an IPO or liquidation event or that the Settlement Master will ever find a buyer for the settlement stake at a commercially reasonable price. That is not disqualifying. For equitybased settlements, uncertainty is inherent. See Uhl v. Thoroughbred Technology & Telecommunications, Inc. , 309 F.3d 978, 982, 986– 87 (7th Cir. 2002) (affirming settlement approving creation of class-owned corporation to be given assets by defendant; acknowledging company will face 'the burdens of any startup company” and that its future “is indeed speculative”). As Yogi Berra may have said, “It’s tough to make predictions, especially about the future.”

Given Clearview's financial circumstances —Judge Andersen attested that Clearview could not “pay any judgment in the tens, never mind the hundreds, of millions of dollars”— the alternative to an uncertain equity stake was not a guarantee of payment but a substantial risk of bankruptcy. That would have left the Class as unsecured creditors even if they won at trial, with only cents on the dollar or perhaps nothing at all.

Objectors also criticize the cash demand option. They write that if Clearview’s revenues are too low, the Settlement Master “might elect to let the option expire in the hope that one of the other three triggering events eventually occurs.” Appellants’ Br. at 29. As Settlement Master, Judge Schenkier is entitled to inspect Clearview’s books, including the amounts set aside from revenue. He also owes a fiduciary duty to the Class. Fiduciary duty standards imposed by ERISA may be instructive here. See Ameritech Benefit Plan Committee v. Communication Workers of America , 220 F.3d 814, 825 (7th Cir. 2000) (comparing fiduciary duties under ERISAPage 15 to those of a common law trustee). An ERISA fiduciary must act 'with the care, skill, prudence, and diligence under the circumstances then prevailing that a prudent man acting in a like capacity and familiar with such matters would use in the conduct of an enterprise of a like character and with like aims.” 29 U.S.C. § 1104(a)(1)(B).

Forgoing the cash demand and banking on a future triggering event might be a risky strategy. Perhaps risking no recovery whatsoever could be justified if it turns out that 23% of Clearview’s valuation greatly exceeds the amount set aside in the revenue accounts and if an IPO or liquidation event appears imminent. Yet perhaps the Class could end up with nothing if, even from that strong financial position, Clearview suddenly crashes and burns. But these possibilities are mere speculation, and any equity-base d settlement involves some risk. See Uhl , 309 F.3d at 987.

Second, and related, the monetary value of the settlement stake is of course uncertain. The district court credited an assertion by the Class, based on confidential financial documents provided by Clearview to Lead Class Counsel and to Judge Andersen, that its valuation as of January 2024 was approximately $225 million, making a 23% stake approximately $51.75 million. Objectors do not dispute that the 23% stake is the product of a vigorous, arm's-length negotiation between the parties, mediated skillfully by Judge Andersen, as the district court found. Further, the estimated dollar amount of the 23% stake is commensurate with other BIPA settlements identified by the court, taking account of differences in defendants' resources and the size of the classes, among other factors. See, e.g., In re Facebook Biometric Information Privacy Litig. , 522 F. Supp. 3d 617 (N.D. Cal. 2021)Page 16 ($650 million), aff'd, 2022 WL 822923 (9th Cir. Mar. 17, 2022); Rivera v. Google LLC , No. 2019-CH-990 (Ill. Cir. Ct. Sept. 28, 2022) ($100 million); In re TikTok, Inc., Consumer Privacy Litig. , 617 F. Supp. 3d 904 (N.D. Ill. 2022) (Lee, J.) ($92 million), appeal dismissed, 2022 WL 19079999 (7th Cir. Oct. 12, 2022); Boone v. Snap Inc. , No. 2022-LA-708 (Ill. Cir. Ct. Nov. 22, 2022) ($35 million).

Of course, Clearview’s valuation is subject to change, but change cuts both ways. If Clearview's valuation balloons, then the settlement stake will be worth more. That feature distinguishes this case from other settlements where uncertain future events—suc h as the amount of forthcoming attorney fee applications—could only reduce the payout, even to zero. See In re Katrina Canal Breaches Litig. , 628 F.3d 185, 196 (5th Cir. 2010) (reversing approval of settlement; noting lack of “any assurance that attorneys’ costs and administrative costs will not cannibalize the entire $21 million settlement”). Here, by contrast, the attorneys will be paid only when the Class is paid, and the Class will be paid more the better Clearview does.

Third, Objectors call that very feature of the settlement “fundamentally unfair” because it “tie[s] the class members’ recovery to the continuation and expansion of the invasion of privacy that the case was brought to stop.” They also claim that settlements structured like this one, with the plaintiff class receiving an equity stake in the defendant, 'present ‘special moral complexity’ because they allow a company that has caused harm to become more profitable, 'all while transforming victims into shareholders.’” Appellants’ Br. at 34-35, quoting Margaret Schaack, Comment, Injury Equity: The Rise of Future Stakes Settlements , 92 Univ. of Chicago L.

Page 17Rev. 1125, 1171 (2025) (criticizing proposed settlement in this case).

Whatever weight these thoughtful attacks might have in other cases, their weight here depends on the potential merits of the unpled and untested constitutional, common-law, and statutory right-to-privacy claims. The Objectors have not identified any statut es that Clearview's current business practices clearly violate. In fact, Clearview has expanded its opt-out program from Illinois to other states (or ceased doing business in other states) when necessary to comply with a new biometric data privacy law, including California and Virginia but not, so far, New York. See California Consumer Privacy Act of 2018 (as amended), Cal. Civ. Code § 1798.100 et seq.; Consumer Data Protection Act, Va. Code § 59.1-575 et seq.; see also Biometric Privacy Act, S.B. 1422A, 2025 –26 Sess. (N.Y. 2025) (passed state senate June 3, 2026). The settlement releases only those claims that 'were, or could have been, advanced' in the MDL, which appears to exclude claims based on future enactments.

Objectors still find Clearview’s business, well, objectionable. They are not alone. The district court expressed concerns with the use and misuse of biometric information, including over-surveillance and racial inequity. On the other hand, the court noted that Clearview’s database has helped solve crimes and exonerate the wrongfully accused. Weighing these considerations and deciding how to respond is the domain of the political branches, and debate continues in Congress and in state legislatures across the country. E.g., Online Privacy Act of 2026, H.R. 8014, 119th Cong. (2026). A court faced with a proposed classaction settlement, by contrast, is limited to deciding whether the settlement is “fair, Page 18reasonable, and adequate.” Fed. R. Civ. P. 23(e). The weaknesses in the substantive claims and other factors supporting the district court's approval put the overall settlement terms within a reasonable range the district court could approve.⁴

We must still note one significant concern here. The district court did not make a finding as to the anticipated value of the cash demand option because it did not examine Clearview’s revenue statements, as Judge Andersen had. On appeal, the Class and Clearview argue that “the only conceivable scenario in which the Class could receive nothing is if the Settlement Master, despite being fully informed, ignores his fiduciary duties to the Class and declines to accept the Cash Demand in favor of a wholly speculative future liquidation event.” Joint Appellees’ Br. at 39–40. If they are right that Judge Schenkier should take the cash demand right before it expires if the equity stake has not yet vested—and they probably are-then the estimated value of the cash demand is important. If the parties come to a substantially similar agreement on remand, the district court itself will need to be informed directly about Clearview's finances and to make appropriate findings regarding the cash demand option.

We now turn from the overall terms of the settlement to the allocation of money under the settlement among the Class members . Objectors' challenge here is at bottom one of procedure. They challenge the lack of a separate class representative, with separate counsel, for the Nationwide Class. To recall, the Class consisted of five groups: state -specific subclasses for California, Illinois, New York, and Virginia, as well as a Nationwide Class encompassing everyone in Clearview’s database (including those who are also members of the subclasses). Under the settlement, each member of the Illinois Subclass receives ten shares of the s ettlement fund, and each member of the California, New York, and Virginia Subclasses receives five shares. Anyone who is a member of only the Nationwide Class receives just one share. Each class representative who asked the district court to approve this s ettlement belonged to both the Nationwide Class and one of the favored statespecific subclasses. Also, no attorney involved in the settlement was responsible for advancing the interests of solely the Nationwide Class, either. Objectors argue the lack of separate representatives and representation for the Nationwide Class requires vacating approval of the settlement. We agree.

Class-action litigation relies on 'structural assurance of fair and adequate representation for the diverse groups and individuals affected.' See Amchem Products, Inc. v. Windsor , 521 U.S. 591, 627 (1997) (affirming appellate court's decision vacating approval of settlement that failed to address important conflicts of interest within huge class). The requirement for approval of a proposed settlement by the class representatives, who owe a fiduciary duty to absent class Page 20members, is one important structural feature that helps provide such assurance. See Eubank v. Pella Corp. , 753 F.3d 718, 719 (7th Cir. 2014).

Not just any class representatives will do. Identifiable groups within the whole class may have materially adverse interests. Conflicts of interest arise, for example, when some class members have already suffered harms while others have only been exposed to a risk of a future injury. See Amchem , 521 U.S. at 626 (claimants already injured by asbestos v. those only exposed to asbestos); Eubank , 753 F.3d at 721 (customers who had already replaced defective windows v. those who had not). Conflicts of interest may also arise, as here, when class members assert different claims with significantly different values or that would entitle them to signifi cantly different forms of relief. See Ortiz v. Fibreboard Corp. , 527 U.S. 815, 857 (1999) (class settlement needed to distinguish between claimants whose injuries were covered by insurance and those whose injuries were not).

Conflicts of interest among identifiable groups within the class, or among recognized subclasses, may be considered in terms of whether “the representative parties will fairly and adequately protect the interests of the class.” Fed. R. Civ. P. 23(a)(4). Co nflict may also be considered in terms of whether a proposed settlement is 'fair, reasonable, and adequate' where that inquiry will ask whether “the class representatives and class counsel have adequately represented the class” and whether “the proposal treats class members equitably relative to each other.” Fed. R. Civ. P. 23(e)(2). Or, when the court certifies a class for settlement, as here, conflicts may be evaluated through both rubrics at the same time. However theorized, Page 21the inquiry is essentially the same: can the court be confident that absent class members have been represented fairly?

The prerequisites for class certification are a good starting point for identifying conflicts of interest. The representative parties may not be able to “fairly and adequately protect the interests of the class' if there are significant differences in the strength of claims of different class members. See Fed. R. Civ. P. 23(a)(2)–( 4). We say 'significant' because minor or merely hypothetical conflicts of interest do not require separate class representatives; the conflict must instead be 'fundamental.' See Dewey v. Volkswagen Aktiengesellschaft , 681 F.3d 170, 184 (3d Cir. 2012) (collecting authorities on “fundamental” standard); Kohen v. Pacific Investment Management Co. , 571 F.3d 672, 680 (7th Cir. 2009) (contrasting 'hypothetical' conflicts with 'real' ones). The cure for a fundamental conflict is 'division into homogeneous subclasses … with separate representation to eliminate conflicting interests of counsel.' Ortiz , 527 U.S. at 856; Dewey , 681 F.3d at 18990. Determining which conflicts are fundamental may at times require a deft touch. See Ortiz , 527 U.S. at 857 (noting “at some point there must be an end to reclassification with separate counsel').⁵

Here, however, we think the need for separate representation should have been apparent from the beginning. Even the original consolidated class-action complaint suggested a wide gulf in the relief the Nationwide Class and each subclass Page 22might have been able to obtain on their own. BIPA authorizes liquidated damages of $1,000 for a negligent violation and $5,000 for a reckless or intentional violation, plus attorney fees and litigation expenses, and it expressly authorizes injunctive relie f. 740 ILCS 14/20(a). Shortly before the filing of this complaint, BIPA served as the basis for the $650 million settlement over Facebook’s use of biometric facial scanning technology, strongly suggesting that the statute could offer substantial relief to the Illinois Subclass here. See Facebook , 522 F. Supp. 3d at 620–21. The California statutory commercial misappropriation claim authorizes liquidated damages of $750, profits attributable to the misappropriation, attorney fees, costs, and punitive damages. Cal. Civ. Code § 3344(a). One of the Virginia claims authorizes punitive damages, too. Va. Code § 8.01-40(A). The Class pled an entitlement to injunctive relief for the California Unfair Competition Law and common law right of publicity claims. The Class did not specifically plead an entitlement to liquidated or punitive damages for the New York claim, though it sought injunctive relief. The only claims pled on behalf of the Nationwide Class were the declaratory judgment and unjust enrichment claims. So, these distinct remedies were the first warning sign.

By the second round of negotiations with Clearview, in any event, Lead Class Counsel should have realized the problem. According to Judge Andersen, the first round of negotiations ended because plaintiffs insisted on a large cash payment, which Clearview maintained it could not afford. By the time the parties returned to the negotiating table, however, he said that “both parties had agreed that any viable class action settlement would need to include the Class receiving a meaningful equity stake in Clearview.' Also according to Judge Andersen, the 23% stake on which the parties eventually Page 23agreed was “particularly the subject of long and contentious negotiation. Too large a percentage ran the risk of preventing Clearview from attracting additional, future investors, which Clearview needed to survive and to thrive; too small a percentage woul d not have provided the Class with a sufficient payout for their claims and their zealous advocacy of those claims.' So, the fight became about just how large a slice of the Clearview pie the Class as a whole could get.

Turning from the overall settlement terms to the question of allocation, however, a common fund like this one can be a warning sign for intraclass conflicts. 'In these zero -sum circumstances, a benefit to one group of class members (in the form of a larger portion of the common fund) comes at the detriment of the other class members (who receive a smaller portion as a result).' Murray v. Grocery Delivery E-Services USA Inc. , 55 F.4th 340, 345 (1st Cir. 2022) (vacating approval of proposed settlement).

The fact that class memberships may overlap is no answer. In Amchem , each of the class representatives “served generally as representative for the whole, not for a separate constituency.' 521 U.S. at 627. In affirming the Third Circuit's decision to vacate approval of a settlement, the Supreme Court quoted the Second Circui t in another asbestos settlement, which summarized the problem nicely:

[W]here differences among members of a class are such that subclasses must be established, we know of no authority that permits a court to approve a settlement without creating subclasses on the basis of consents by members of a unitary class, some of whom happen to be members of the distinct subgroups. The class representa-

Page 24tives may well have thought that the Settlement serves the aggregate interests of the entire class. But the adversity among subgroups requires that the members of each subgroup cannot be bound to a settlement except by consents given by those who understand that their role is to represent solely the members of their respective subgroups.

Id. (alteration in original), quoting In re Joint Eastern & Southern District Asbestos Litig. , 982 F.2d 721, 743 (2d Cir. 1992), modified on reh'g sub nom. In re Findley , 993 F.2d 7 (2d Cir. 1993). This case presents an inverse problem: none of the named class representatives was in a position to represent solely the interests of the Nationwide Class in allocating the settlement.

Judge Andersen's affidavit did not say how the allocation was determined. The district court wrote that “the pro rata distribution of shares across each subclass … was not the product of shadowy, backroom negotiations between Plaintiffs and Defendants but reflects the strength and viability of the respective … claims.” The district court’s comment about the differences among the strength and viability of claims under different state laws may well turn out to be substantively correct, at least in terms of th e rank order of the different state -law claims. BIPA is stronger than the other statespecific claims, which are stronger than the unjust enrichment claim.

But it is difficult to accept any particular set of ratios as fair when those who would get the short end of the stickmembers of only the Nationwide Class—were not represented separately in the negotiations and did not themselves agree to that short end. As the Second Circuit Page 25explained in a similar case involving three classes of copyright claims based on copyright registration status and date, Categories A, B, and C, where the class representatives all held multiple classes of claims but none held only Category C claims:

Although all affected members of the plaintiff class are interested in maximizing their individual compensation, severally they accomplish that goal in different ways. To authors who own works in all three categories, how their compensation is allotted amo ng their claims is irrelevant; what matters is the bottom line.…

We know that Category C claims are worth less than the registered claims, but not by how much. Nor can we know this, in the absence of independent representation.

In re Literary Works in Electronic Databases Copyright Litig. , 654 F.3d 242, 246, 251, 253 (2d Cir. 2011) (vacating approval of settlement).

Assuming the class representatives and their attorneys negotiated the ratios among themselves, we are not saying that the result was unreasonable or unfair. Rather, the problem is that the settlement lacks adequate structural protections for us to tell either way. Appointment of separately counseled class representatives for identifiable, significantly different groups of claimants with fundamentally conflicting interests is Rule 23's primary mechanism for such protection. Only then can the district court p roperly fulfill its role in scrutinizing the settlement. See Murray , 55 F.4th at 351 ('[E]nsuring that claims marked by significantly different elements or de- Page 26fenses receive appropriate relative weight in a class settlement should be done in the first instance through negotiations between counseled representatives of the different groups of class members, not by the district court, unless the appropriate relative weight is clear-cut.').⁶

Finally, although we need not decide how heavily to weigh the substitution of all eight original class representatives after each refused to agree to the settlement, we agree with Objectors that the wholesale substitution was a potential 'red flag.' See Ap pellants’ Br. at 40, quoting Eubank , 753 F.3d at 721, 723–24. The district court noted the replacement but did not otherwise discuss it in evaluating the adequacy of representation. If another settlement is reached on remand , the district court should consider the substitution and the reasons for it in evaluating whether the class representatives have adequately protected the interests of the class, as opposed to the attorneys' interest in settling this litigation.

We VACATE the district court’s approval of the settlement and REMAND for further proceedings consistent with this opinion. Costs to Objectors. Fed. R. App. P. 39(a).