Khatibi v. Hawkins

FOR PUBLICATION UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT AZADEH KHATIBI, M.D., an No. 24-3108 individual; DO NO HARM, a D.C. No. Virginia nonprofit corporation, 2:23-cv-06195- MRA-E Plaintiffs - Appellants, Central District of California, v. Los Angeles RANDY HAWKINS, in his official ORDER capacity as President of the Medical Board of California; LAURIE ROSE LUBIANO, in her official capacity as Vice President of the Medical Board of California; REJI VARGHESE, in his official capacity as Executive Director of the Medical Board of California; MARINA O'CONNOR, in her official capacity as Chief of Licensing, Medical Board of California; RYAN BROOKS, in his official capacity as Secretary of the Medical Board of California, Defendants - Appellees. Filed December 29, 2025 2 KHATIBI V. HAWKINS Before: A. Wallace Tashima, Jacqueline H. Nguyen, and Salvador Mendoza, Jr., Circuit Judges. Order; Dissent by JudgeVanDyke; Dissent by Judge Tung SUMMARY * First Amendment/Government Speech The panel denied a petition for panel rehearing and rehearing en banc of the panel’s decision affirming the district court’s dismissal of an action, brought by a physician instructor of continuing medical education (CME) courses and a nonprofit comprised of healthcare professionals and policymakers, alleging that the Medical Board of California’s requirement that CME courses eligible for credit include information about implicit bias violates the Free Speech Clause of the First Amendment. Dissenting from the denial of rehearing en banc, Judge VanDyke, joined by Judges Bumatay and Tung, wrote that the panel erred in concluding that CME courses are government speech devoid of any First Amendment protection because (1) California has not historically used CME courses to communicate the state’s own messages, (2) those attending CME courses would be unlikely to perceive the instructor’s message as the government’s * This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.  KHATIBI V. HAWKINS 3 message, and (3) the state’s regulations otherwise exert very little control over CME instructors’ messages. The panel’s decision puts this circuit out of step with the precedent of the Supreme Court and sister circuits, and even this circuit’s precedent. Dissenting from the denial of rehearing en banc, Judge Tung, joined by Judges Bumatay and VanDyke, wrote that private instructors of continuing medical education courses do not engage in “government speech,” for the simple reason that they are not the government and they do not speak for the government. A law requiring them to convey a viewpoint they find objectionable thus restricts their private expression and is not exempt from First Amendment scrutiny. ORDER The panel unanimously voted to deny the petition for panel rehearing. Judges Nguyen and Mendoza voted to deny the petition for rehearing en banc and Judge Tashima so recommended. The full court was advised of the petition for rehearing en banc. A judge requested a vote on whether to rehear the matter en banc. The matter failed to receive a majority of the votes of the nonrecused active judges in favor of en banc consideration. Fed. R. App. P. 40. The petitions for panel rehearing and rehearing en banc (Dkt. No. 64) are DENIED. 4 KHATIBI V. HAWKINS VANDYKE, Circuit Judge, joined by BUMATAY and TUNG, Circuit Judges, dissenting from the denial of rehearing en banc In a published opinion, a panel of our court held that a medical course taught by a private instructor and accredited by private entities is government speech unprotected by the First Amendment. That conclusion isn’t merely incorrect— it puts our circuit out of step with Supreme Court precedent, our sister circuits’ precedent, and even our own precedent. Our court’s denial of rehearing en banc passes up the opportunity to rectify the panel’s mistaken conclusion. I respectfully dissent from that denial. The state of California conditions medical licenses on the completion of accredited continuing medical education (“CME”) classes. Cal. Code Regs. tit. 16, § 1336(a). Plaintiff Dr. Azadeh Khatibi and at least one member of plaintiff Do No Harm, a nonprofit corporation, create content for those CME courses—content that involves speech protected by the First Amendment. But under California’s relatively recent changes to its CME regulations, Plaintiffs must affirm controversial state-sanctioned political and ideological beliefs by incorporating into any CME course “curriculum that includes the understanding of implicit bias.” Cal. Bus. & Prof. Code § 2190.1(d)(1). Plaintiffs challenged that requirement in federal court, arguing that CME courses are private speech and that the government cannot constitutionally compel CME instructors to adopt the state’s views on divisive political subjects. The district court dismissed Plaintiffs’ claims, finding that the content of CME courses—despite being crafted and taught by private parties—constitutes government speech. The panel affirmed.  KHATIBI V. HAWKINS 5 The panel misread the Supreme Court’s government speech precedent, deviating from the Supreme Court’s instructions and creating a split with our sister circuits by focusing on the scope of California’s regulation of CME courses rather than examining the manner in which California’s regulations shape or convey the messages involved in CME instruction. The result: an expansive government speech doctrine that discards the Supreme Court’s cautionary instruction in Matal v. Tam, 582 U.S. 218, 235 (2017). A proper analysis—as prescribed by the Supreme Court, our own court’s prior cases, and our sister circuits—reveals that California’s prior CME regulations did not meaningfully express or shape messages through CME courses. Because California has not historically used CME courses to communicate the state’s own messages, because those attending CME courses would be unlikely to perceive the instructor’s message as the government’s, and because the state’s regulations otherwise exert very little control over CME instructors’ messages, the panel erred in concluding that CME courses are government speech devoid of any First Amendment protection. I. Similar to continuing legal education requirements for lawyers, CME course attendance is required for California physicians and surgeons wishing to maintain their licenses. Cal. Code Regs. tit. 16, § 1336. California’s express goal for having these CME regulations is to “ensure the continuing competence of licensed physicians and surgeons,” and to that end the Medical Board of California is authorized to “adopt and administer standards for the continuing education of … licensees.” Cal. Bus. & Prof. Code § 2190. At issue here is California’s statutory requirement, added in 2019, that a CME instructor must include in his or her course 6 KHATIBI V. HAWKINS “curriculum that includes the understanding of implicit bias.” Cal. Bus. & Prof. Code § 2190.1(d)(1). An instructor can satisfy that requirement by providing “[e]xamples of how implicit bias affects perceptions and treatment decisions,” or by detailing “strategies to address how unintended biases in decisionmaking may contribute to health care disparities.” Id. § 2190.1(e). Other standards broadly dictate that the content of CME courses must—unsurprisingly—be medical in nature. Cal. Bus. & Prof. Code § 2190.1(a); Cal. Code Regs. tit. 16, § 1337.5(a)(3). CME lectures must “maintain, develop, or increase the knowledge, skills, and professional performance that a physician and surgeon uses to provide care,” which may be done by educational activities that “include, but are not limited to” a broad list of options such as “scientific or clinical content with a direct bearing on the quality or cost-effective provision of patient care, community or public health, or preventive medicine.” Cal. Bus. & Prof. Code § 2190.1(a). California’s regulations echo these same options. Cal. Code Regs. tit. 16, § 1337.5(a)(3). More granular subject matter requirements are imposed on CME attendees: physicians with a certain percentage of elderly patients must complete CME courses in geriatric medicine, and all physicians and surgeons must complete CME courses on either pain management and the treatment of terminally ill patients or on the treatment of opiate-dependent patients. Cal. Bus. & Prof. Code §§ 2190.3, 2190.5, 2190.6. California also allows CME attendees to earn at most 30% of the required credits from CMEs on medical office management. Id. § 2190.15. Other standards lay out requirements for CME instructors, including that the need for the course be “maintained on file,” and a handful of other miscellaneous requirements  KHATIBI V. HAWKINS 7 unrelated to CME content. Cal. Code Regs. tit. 16, § 1337.5(a)(1)–(2), (4)–(7). CME courses must also “contain curriculum that includes cultural and linguistic competency in the practice of medicine.” Cal. Bus. & Prof. Code § 2190.1(b). The Code directs the private associations that accredit CME courses to develop standards to assess compliance with the cultural and linguistic competency requirement and mandates that CME courses “shall address at least one or a combination of” a long list of possible ways to include such content. Id. §§ 2190.1(b)(3), 2190.1(c). California does “not give prior approval to individual courses or programs.” Cal. Code Regs. tit. 16, § 1337.5(b). Instead, California outsources the accreditation of CME courses to private entities. Cal. Bus. & Prof. Code § 2190.1(g); Cal. Code Regs. tit. 16, § 1337(a). The only involvement California maintains in the process is that state actors can “randomly audit courses or programs submitted for credit in addition to any course or program for which a complaint is received.” Cal. Code Regs. tit. 16, § 1337.5(b). When a course is audited, “course organizers will be asked to submit” certain information, like the rationale for the course, the course content, educational objectives, attendance records, and the like. Id. Dr. Khatibi and at least one member of Do No Harm have taught CME courses in California. They object to California’s new implicit bias requirement, Cal. Bus. & Prof. Code § 2190.1(d), on the ground that the regulation requires them to express the state’s controversial viewpoint in CME courses that they created and compiled on their own and that were approved by private CME accreditors. The district court dismissed Plaintiffs’ complaint, finding that teaching CME courses is government speech and that California was 8 KHATIBI V. HAWKINS therefore free to require CME instructors to express the state’s view about implicit bias. The panel affirmed. II. When the government speaks for itself, the First Amendment’s protections for private expression are not implicated. Pleasant Grove City v. Summum, 555 U.S. 460, 467–68 (2009). But sometimes the line between government and private speech blurs, necessitating a more detailed analysis to assess “whether the government intends to speak for itself or to regulate private expression.” Shurtleff v. City of Boston, 596 U.S. 243, 252 (2022). The Supreme Court has highlighted several factors of particular importance to that analysis: “the history of the expression at issue; the public’s likely perception as to who (the government or a private person) is speaking; and the extent to which the government has actively shaped or controlled the expression.” Id. Because of the risk that “government could silence or muffle the expression of disfavored viewpoints” by characterizing that expression as government speech and therefore outside the First Amendment’s ambit, the Supreme Court has admonished courts to exercise “great caution” before extending the government speech doctrine. Matal, 582 U.S. at 235. In line with this warning, the Supreme Court has deemed that “simply affixing a government seal of approval” to private speech fails to transform private speech into government speech. Id. A. While the panel in this case paid lip service to the three- factor Shurtleff test, its approach boils down to a single- factor analysis: does the government “heavily” or “actively” regulate CMEs? The panel’s application of the Shurtleff test can be aptly summarized: Does the history of the expression  KHATIBI V. HAWKINS 9 indicate that CME courses are government speech? Yes, because of the broad scope of the government’s historical regulation of CMEs. Does the public likely perceive the government as speaking? Yes, because of the broad scope of the government’s CME regulations. Does the government exercise sufficient control over the message expressed in CME courses? Yes, because of the broad scope of the government’s CME regulations. This one-factor-to-rule- them-all test is not the test that prior cases have prescribed. The Supreme Court’s precedent, our own prior cases, and cases from our sister circuits all indicate that the analysis is not so simple—that the mere fact of extensive regulation is far from the be-all end-all conclusion of the analysis. Instead, prior cases establish that the analysis focuses not on the mere scope of the state’s regulations, but instead on the government’s particular involvement in shaping the message being expressed. Under Shurtleff, it is obvious that merely regulating private speech does not necessarily transform that speech into government speech: “we conduct a holistic inquiry designed to determine whether the government intends to speak for itself or to regulate private expression.” Shurtleff, 596 U.S. at 252. Shurtleff’s subsequent description of the government speech analysis factors further clarifies that the government’s involvement in actually expressing or shaping the message is the touchstone of the analysis: “the history of the expression at issue; the public’s likely perception as to who (the government or a private person) is speaking; and the extent to which the government has actively shaped or controlled the expression.” Id. at 252 (emphases added); see also Matal, 582 U.S. at 238 (describing the analysis in Walker v. Texas Div., Sons of Confederate Veterans, Inc., 576 U.S. 200 (2015), as partially dependent on whether the State 10 KHATIBI V. HAWKINS historically used license plates “to convey state messages” and whether the state maintained “direct control over the messages conveyed” (emphases added) (quoting Walker, 576 U.S. at 213)). While our own precedent on the government speech doctrine is fairly sparse, our cases that consider the doctrine display a similar emphasis on the government’s involvement with the message being conveyed. For instance, we have considered whether the “message is ‘from beginning to end’ that of the State.” Delano Farms Co. v. Cal. Table Grape Comm’n, 586 F.3d 1219, 1228 (9th Cir. 2009) (emphasis added) (quoting Paramount Land Co. LP v. Cal. Pistachio Comm’n, 491 F.3d 1003, 1012 (9th Cir. 2007)); see also Ranchers Cattlemen Action Legal Fund United Stockgrowers of Am. v. Vilsack, 6 F.4th 983, 990 (9th Cir. 2021) (same). These cases suggest that if a given regulation does not meaningfully control the message being communicated, mere regulation—even extensive regulation—carries little weight, if any, in proving that the government is actually speaking for itself. Cases from other circuits likewise focus on whether the government is articulating its own message both overall and for each Shurtleff factor, rather than on the mere breadth of the government’s regulations. See Sutliffe v. Epping Sch. Dist., 584 F.3d 314, 331 (1st Cir. 2009) (finding government speech where a town “communicated an important message about itself”); Women for Am. First v. Adams, No. 21-485- cv, 2022 WL 1714896, at *4 (2d Cir. May 27, 2022) (finding that certain murals were government speech because the City propagated “its own message” through the murals); Brown v. Yost, 133 F.4th 725, 734 (6th Cir. 2025) (determining that summaries of proposed ballot initiatives were not government speech because they did not “historically  KHATIBI V. HAWKINS 11 convey[] government messages”); Wandering Dago, Inc. v. Destito, 879 F.3d 20, 36 (2d Cir. 2018) (finding, as part of the government speech analysis, that a private party’s participation in a government program would likely not “be viewed by the public” as the government adopting the private party’s speech as its own); Little v. Llano Cnty., 138 F.4th 834, 860 (5th Cir. 2025) (en banc) (emphasizing that the third Shurtleff factor considers “the extent to which the government has actively shaped or controlled the expression” (quoting Shurtleff, 596 U.S. at 252)). Because in conducting the three-factor Shurtleff analysis the panel here diverged from the guidance of the Supreme Court, our own precedent, and the reasoning of our sister circuits, the panel’s analysis of each factor was improperly skewed by its heavy reliance on the existence of numerous regulations that have little to no connection to shaping the messages conveyed in CME courses. An analysis that properly focuses on the formation of those messages reveals that California has historically had almost no involvement with the content of the speech expressed by CME instructors: not historically, nor in the eyes of the public, nor in terms of controlling CME expression. B. Historically, California has rarely, if ever, communicated government messages through CME courses. This is largely because most of California’s CME regulations have not required that CME courses include specific content. Begin with California’s “content 12 KHATIBI V. HAWKINS standards” in section 2190.1(a), which suggest that qualified CMEs may include, but are not limited to, educational activities that meet any of the following criteria: (1) Have a scientific or clinical content with a direct bearing on the quality or cost- effective provision of patient care, community or public health, or preventive medicine. (2) Concern quality assurance or improvement, risk management, health facility standards, or the legal aspects of clinical medicine. (3) Concern bioethics or professional ethics. (4) Are designed to improve the physician- patient relationship and quality of physician- patient communication. Cal. Bus. & Prof. Code § 2190.1(a) (emphasis added). Other than unsurprisingly assuming that CME courses will be broadly related to the medical field, this incredibly expansive, nonbinding, and nonexhaustive list imposes functionally no content restriction on CME instructors, giving them free rein to pick their topics and decide what to say about them. If someone sets out in good faith to teach a CME course, it’s hard to imagine what topic would be excluded under this provision. The provision is so broad that it does not even prevent CME instructors from expressing messages that conflict with other courses, or indeed even presenting conflicting viewpoints within the same course.  KHATIBI V. HAWKINS 13 The content standards in California’s regulations merely parrot the language from section 2190.1(a): “The content of the course or program shall be directly related to patient care, community health or public health, preventive medicine, quality assurance or improvement, risk management, health facility standards, the legal aspects of clinical medicine, bioethics, professional ethics, or improvement of the physician-patient relationship.” Cal. Code Regs. tit 16, § 1337.5(a)(3). As such, this second content standard— essentially the same as the first, but now with mandatory language—adds nothing to the analysis. Unlike the monuments that the Supreme Court deemed to be government speech in Summum, California’s generic content standards do not aim “to convey some thought or instill some feeling.” Summum, 555 U.S. at 470. Similarly, unlike the license plates deemed government speech in Walker, California’s CME content standards come nowhere close to communicating messages as specific as “a graphic” or “a slogan.” Walker, 576 U.S. at 211. Properly assessed for any historically communicated government message, California’s content standards do nothing to show that the government has historically used CME courses to communicate a government message. California’s regulations include “Criteria for Acceptability of Courses” that lay out generic quality standards for CME courses: standards for who can be certified as a CME instructor, requirements that course rationales, objectives, and teaching methods be kept on file, and the like. Cal. Code Regs. tit. 16, § 1337.5(a)(1)–(2), (4)– (7). These quality requirements evince the same absence of a historical government message as the content standards: they do not require that any specific message be conveyed during CME courses and therefore do nothing to 14 KHATIBI V. HAWKINS demonstrate that the messages historically communicated through CME courses were government messages as opposed to the messages of private CME instructors. The state’s oversight role beyond its written regulations does nothing to move the needle closer to government speech. California’s CME regulations allow the government to audit individual courses, which it apparently occasionally does. But because California’s content and quality standards do not show that the state has historically conveyed messages through CME courses, neither does the state’s ability to audit CME courses. Whether the potential for an audit means that CME courses express a government message depends entirely on what the state is auditing for: auditing for compliance with content and quality standards that don’t require any particular government message means the audits cannot have required historical government messages either. And outside of sporadic audits, the state has outsourced the bulk of the approval work to private accreditors, making it even less likely that any state message has somehow been historically communicated through CME courses. Cal. Bus. & Prof. Code § 2190.1(g); Cal. Code Regs. tit. 16, § 1337.5(b). Finally, California’s requirements that certain licensees take CME courses covering specific subject matter (like geriatric medicine) or capping how much credit can be earned from courses covering specific subject matter (like medical office management) are restrictions only on CME attendees, not on the private speakers involved in the CME process: the CME instructors. Cal. Bus. & Prof. Code §§ 2190.3, 2190.5, 2190.6, 2190.15. Nowhere does California force private parties to create CME courses on those specific topics. As such, those requirements also fail  KHATIBI V. HAWKINS 15 to evidence a history of California speaking through CME courses. The panel concluded otherwise, but the closest it got to identifying any specific government message historically communicated using CME courses is California’s requirement that CME instructors discuss cultural and linguistic competence. Cal. Bus. & Prof. Code § 2190.1(b)– (c). But as discussed below in applying the third Shurtleff factor, even the cultural and linguistic competence provisions are so broad and leave so much discretion to CME instructors that they practically exert essentially no control over the messages conveyed by CME courses, which in turn makes it impossible to point to any discrete state message communicated through CME courses as a result of these provisions. To its credit, the panel did attempt (albeit unsuccessfully) to distill from California’s historical regulations one “overarching” government message: “what is necessary to ensure the continuing competence of licensed physicians” that “reflects the State’s evolving judgment of what subjects it has deemed essential” for doctors to know. California argued in the same vein that its regulations “convey[] a message regarding what subjects the State deems important for doctors to know.” First off, this amorphous, super-generalized message is communicated, at most, only by the existence of the regulations, not by the content of any individual CME courses. This looks nothing like the government speech in Summum and Walker, in which the government used individual monuments to communicate discrete messages and individual license plates to communicate specific graphics, slogans, and the like. See Summum, 555 U.S. at 16 KHATIBI V. HAWKINS 470; Walker, 576 U.S. at 211. The analysis in Summum and Walker suggests that Shurtleff’s reference to “the history of the expression at issue” looks not to whatever broad governmental priorities can be inferred from the mere fact of regulation, but to the discrete, particular messages the government has historically communicated using the regulated medium. Shurtleff, 596 U.S. at 252 (drawing the “history of the expression” factor from Summum and Walker). This is confirmed by Matal, in which the Court determined that trademarks were not government speech despite being significantly regulated. Matal, 582 U.S. at 239. Applying the panel’s logic, the Court could have inferred from the heavy regulation of trademarks that the government was communicating its “evolving judgment” about the importance of trademark protections. But that is not how the Court approached its analysis. Instead, it flatly stated that “[t]rademarks have not traditionally been used to convey a Government message.” Matal, 582 U.S. at 238. Inferring a message from just the mere fact of extensive regulation and treating that nebulous message as capable of satisfying the first Shurtleff element would effectively erase the historical element altogether, an outcome hard to square with the Supreme Court’s treatment of the historical element as a discrete and meaningful piece of the analysis. But second, even if we ignore that the panel’s reasoning clashes with Supreme Court precedent and functionally rewrites the test prescribed by Shurtleff, the “overarching message” here, as articulated by the panel and by California, is at such a high level and is so nonspecific that it communicates no more than “simply affixing a government seal of approval.” Matal, 582 U.S. at 235. If all that it takes to transform private speech into government speech is the government’s implied signal that it thinks some number of  KHATIBI V. HAWKINS 17 topics are important, or an implied “evolving judgment of what subjects it has deemed essential” for people to know, then a mere government seal of approval (which is essentially the same thing) would suffice to swallow private speech. Yet Matal indicates the exact opposite: that “simply affixing a government seal of approval” is categorically insufficient to transform private speech into government speech, notwithstanding the high-level messages inferable from such a seal. Id. The first Shurtleff factor therefore weighs in favor of CME messages being private speech, and the panel erred in concluding otherwise. C. Based purely on the fact that California “heavily” regulates CME courses, the panel concluded that “‘common sense’ commands that licensees could attribute approved CMEs’ content to California.” panel reached this conclusion despite acknowledging that Dr. Khatibi “plausibly alleged facts suggesting that [CME] attendees treat her as the person responsible for CME content.” First, this analysis runs directly counter to “accept[ing] as true all well-pleaded factual allegations, and constru[ing] all factual inferences in the light most favorable to the plaintiff,” which we are required to do at the motion-to- dismiss stage. Parents for Priv. v. Barr, 949 F.3d 1210, 1221 (9th Cir. 2020). As the panel recognized, Dr. Khatibi clearly alleged that the attendees of her CME courses view her content as her own, not the government’s. That allegation, irrespective of California’s regulatory scheme, suggests that the public would likely believe that a private person is speaking through CME courses. See Shurtleff, 596 U.S. at 252. At the motion-to-dismiss stage, that should be the end of the analysis of Shurtleff’s public perception factor. 18 KHATIBI V. HAWKINS The panel, however, went on to draw inferences unfavorable to Plaintiffs based on their allegations. The basis for these unfavorable inferences, in the panel’s eyes, is the Supreme Court’s language from Ashcroft v. Iqbal that “[d]etermining whether a complaint states a plausible claim for relief … requires the reviewing court to draw on its judicial experience and common sense.” 556 U.S. 662, 679 (2009) (citation omitted). The subsequent paragraphs in that decision, however, suggest that the point of resorting to common sense is to weed out “legal conclusion[s]” and to ensure that a plaintiff’s allegations are logically consistent, not to flatly contradict a well-pleaded allegation the court is supposed to accept as true. Id. at 679–80. And even assuming it can be permissible to rely on “judicial experience and common sense” as the panel suggests, the most obvious cognate to CME from our own “judicial experience” would obviously be the comparable continuing legal education (“CLE”) requirements of our profession, which like the medical profession is heavily regulated, often in very similar ways. See, e.g., Cal. Bus. & Prof. Code § 6070.5 (directing the State Bar to adopt regulations requiring “that the mandatory continuing legal education … curriculum for all licensees … includes training on implicit bias and the promotion of bias-reducing strategies”). Every judge has sat through hours upon hours of CLE, and surely that experience and common sense support the eminent plausibility of Plaintiff’s allegation that CLE attendees perceive CLE instructors as speaking on their own behalf, not on behalf of the government. Given the similarity between CME and CLE requirements, there is little reason to believe that CME attendees would think any differently—certainly nothing in our “judicial experience” compels a different conclusion. Just like the first factor, the  KHATIBI V. HAWKINS 19 second Shurtleff factor also supports a finding that CME courses are private speech. 1 D. The panel’s analysis of the third Shurtleff factor involved a lengthy recitation of the laws that California has in place. As discussed above, the mere existence of a plethora of regulations is not enough to prove that the government is speaking. We must instead consider the ways in which the regulations “actively shape[] or control[] the expression” involved in the CME lectures. Shurtleff, 596 U.S. at 252; cf. id. at 256 (“[I]t is Boston’s control over the flags’ content and meaning that here is key; that type of control would indicate that Boston meant to convey the flags’ messages.”). Conducting that analysis reveals that the state in fact exercises very little control over CME expression, because every part of the regulations that have to do with the content of the CME courses is so broad as to exercise practically no control over the courses’ messages. For the same reasons that many of California’s regulations do not evidence a history of the state speaking through CME courses, the same regulations impose no meaningful government control over the expression involved with CME courses. The breadth of California’s content standards leaves CME instructors free to speak about whatever medical topics they choose. See Cal. Bus. & Prof. 1 The panel contended that “[e]ven assuming that the public perception factor favors Dr. Khatibi, [the] ultimate conclusion would remain the same” because the “remaining factors of history and extent of state control decisively” favor California. But this conclusion relies on the panel’s erroneous analysis of the history and extent-of-control factors. As explained, properly analyzed, neither of those factors favor the government—much less “decisively” so. 20 KHATIBI V. HAWKINS Code § 2190.1(a); Cal. Code Regs. tit. 16, § 1337.5(a)(3). Within the factual context of Walker, this would be akin to the government saying, “Create whatever license plate design you want, as long as it’s a license plate.” Such a permissive content requirement, a far cry from Texas’s granular “sole control over the design, typeface, color, and alphanumeric pattern for all license plates,” would certainly not have led the Court to conclude that “Texas maintains direct control over the messages conveyed on its specialty plates.” Walker, 576 U.S. at 213. Given that Walker “likely marks the outer bounds of the government-speech doctrine,” Matal, 582 U.S. at 238, this vast disparity between Texas’s control over the license plates in Walker and California’s exercise of control over the messages conveyed in CME courses means that the third Shurtleff factor also favors Plaintiffs. Since California’s miscellaneous quality standards do not impact the messages conveyed in CME courses, those regulations cannot be characterized as “actively shap[ing] or controll[ing] the expression” at issue. 2 Shurtleff, 596 U.S. at 252; see Cal. Code Regs. tit. 16, § 1337.5(a)(1)–(2), (4)–(7). The subject matter requirements for certain licensees in 2 The panel makes much of the fact that “Dr. Khatibi admits that her courses have complied with all CME requirements,” which the panel reads as Dr. Khatibi “conced[ing] that her CMEs have been shaped by California.” This is another example of the panel putting dispositive weight on the mere fact of regulation, rather than properly analyzing whether those regulations shape the messages conveyed by CME instructors. All regulations require compliance, regardless of whether they control or shape speech. If regulatory compliance alone is sufficient to resolve the analysis of the third Shurtleff factor, then the distinction drawn by the Court between “the government intend[ing] to speak for itself” and the government merely “regulat[ing] private expression” would be meaningless. Shurtleff, 596 U.S. 243, 252 (2022).  KHATIBI V. HAWKINS 21 California Business and Professions Code sections 2190.3, 2190.5, 2190.6, and 2190.15 are imposed on CME attendees, not CME instructors, and therefore likewise fail to provide examples of the state asserting control over the messages CME instructors convey. As part of its control analysis, the panel again leaned on California’s cultural and linguistic competency requirements. Cal. Bus. & Prof. Code § 2190.1(c). The panel cited two pages’ worth of the associated statute in an attempt to demonstrate the extent of the control California exerts over CME instructors’ speech. If the cited statute listed requirements that all had to be met, the panel’s argument might be stronger. But the statute is disjunctive: CME lectures must address “at least one or a combination of” the subsequently listed items. Cal. Bus. & Prof. Code § 2190.1(c). As a result, the length of the statute actually serves to undercut the panel’s point: because CME instructors can choose among almost unlimited options to satisfy the cultural and linguistic competence requirement, section 2190.1(c) in fact exercises very little control over CME instructors’ speech. This provision doesn’t merely leave “the development of the remaining details” to the private speakers who create CME courses. Johanns v. Livestock Mktg. Ass’n, 544 U.S. 550, 561 (2005). Giving CME instructors a choice between a whole host of options for one small aspect of CME lectures and wholesale control over the rest of the message means that CME instructors are functionally building their courses from the ground up, with no meaningful state-imposed guardrails or direction. Because neither the cultural and linguistic competence provision nor the generic content requirements significantly control or limit CME instructors’ messages, the third 22 KHATIBI V. HAWKINS Shurtleff factor too weighs against finding that the government is speaking through CME courses. III. In California, CME courses are created, prepared, approved, and accredited by private actors. While, as one might expect, the state extensively regulates CME courses, it has not historically used that regulation to control the courses’ messages. Nor is the mere scope of California’s regulatory scheme a good reason to conclude, at the motion- to-dismiss stage and in the face of Plaintiffs’ well-pleaded allegations, that CME attendees perceive instructors as relaying the government’s views. As a factual matter, and certainly as plausibly pled in this case, they don’t. And the breadth of California’s CME regulations generally belies the fact that the state actually exercises very little control over the messages expressed by CME instructors. The panel’s conclusion in this case that the mere breadth of CME regulation inherently expresses a governmental message about California’s priorities improperly rewrites the test prescribed by the Supreme Court in Shurtleff. Our court should have reheard this case en banc to correct that improperly anemic governmental speech analysis and to prevent the government from so easily coopting private speech.  KHATIBI V. HAWKINS 23 TUNG, Circuit Judge, joined by BUMATAY and VANDYKE, Circuit Judges, dissenting from the denial of rehearing en banc: This case is about whether private instructors of continuing medical education courses engage in “government speech” when a State requires them (over their objection) to teach that “implicit bias” “lead[s] to disparities in health outcomes.” The answer is no. Such private instructors do not engage in “government speech,” for the simple reason that they are not the government and they do not speak for the government. A law requiring them to convey a viewpoint they find objectionable thus restricts their private expression and is not exempt from First Amendment scrutiny. Because the panel concluded otherwise, I respectfully dissent from the denial of rehearing en banc. * * * The concept of “implicit bias” is controversial. Popularized in the 2000s, it espouses the view that individuals harbor unconscious biases that favor whites and males over blacks and females (or other groups) that have resulted in disparities in income, job opportunities, and most relevant here, healthcare outcomes. See, e.g., Assemb. B. 241, 2019–2020 Leg., Reg. Sess. (Cal. 2019) (enacted) (codified at Cal. Bus. & Prof. Code §§ 2190.1(d)–(g), 2736.5(a)–(c), 3524.5(a)–(d)). Plaintiffs here—a female physician and a nonprofit—dispute the validity of the concept and its relevance to their curriculum. ER 36–37, 39. “Implicit bias” theory, they say, is rooted in neither evidence nor fact, disregards other potential factors that could better explain outcome disparities, and hastily (and inaccurately) identifies racism or sexism as the primary cause. Id. at 35– 24 KHATIBI V. HAWKINS 37. In Plaintiffs’ view, the theory is also divisive, producing resentment by needlessly setting one racial (or other) group against another. Id. at 35, 39. For these reasons, Plaintiffs object to teaching “implicit bias” in their continuing medical education courses. Id. at 36–39. The State of California has taken a firm side in this debate. It has sought to disseminate the idea of “implicit bias” through various mandatory training programs. To that end, and for our purposes here, the California legislature in 2019 enacted a statute requiring that private instructors teach “implicit bias” in all their continuing medical education courses relating to “direct patient care.” Cal. Bus. & Prof. Code § 2190.1(d). In that Act, the legislature “declares” “find[ings]” that purportedly support its mandate—though the Act itself does not cite evidence for those findings: • Implicit bias . . . exists, and often contributes to unequal treatment of people based on race, ethnicity, gender identity, sexual orientation, age, disability, and other characteristics[.] • Implicit bias contributes to health disparities by affecting the behavior of physicians and surgeons, nurses, physician assistants, and other healing arts licenses[.] • African American women are three to four times more likely than white women to die from pregnancy-related causes nationwide. • African American patients often are prescribed less pain medication than  KHATIBI V. HAWKINS 25 white patients who present the same complaints[.] • African American patients with signs of heart problems are not referred for advanced cardiovascular procedures as often as white patients with the same symptoms[.] • Implicit gender bias also impacts treatment decisions and outcomes. Women are less likely to survive a heart attack when they are treated by a male physician and surgeon. Assemb. B. 241, 2019–2020 Leg., Reg. Sess. § 1 (Cal. 2019). In accordance with these “findings,” the Act states that “all continuing medical education courses shall contain curriculum that includes the understanding of implicit bias.” Cal. Bus. & Prof. Code § 2190.1(d)(1). Specifically, the Act requires private instructors to incorporate “at least one or a combination” of the following in virtually all courses: (1) Examples of how implicit bias affects perceptions and treatment decisions of physicians and surgeons, leading to disparities in health outcomes. (2) Strategies to address how unintended biases in decisionmaking may contribute to health care disparities by shaping behavior and producing differences in medical treatment along lines of race, ethnicity, gender identity, sexual 26 KHATIBI V. HAWKINS orientation, age, socioeconomic status, or other characteristics. Id. § 2190.1(e). * * * Let us be clear about a few points. First, no one disputes that Plaintiffs are a private instructor and a private entity who teach courses to physicians so that the physicians can maintain their licenses. Plaintiffs are not government employees or government agents; nor are they funded by the government. See Khatibi v. Hawkins, 145 F.4th 1139, 1151 (9th Cir. 2025) (“[T]he State certainly expects, if not relies[] on[,] the participation of private parties in executing the CME scheme”); Cal. Code Regs. tit. 16, §§ 1337, 1337.5 (expressly opening up continuing education courses to approved private providers). Second, no one contends that the government owns (by copyright or otherwise) the course materials that Plaintiffs put together. It is not disputed that the course materials (and the content contained therein) belong to Plaintiffs. Third, the law at issue does not purport to transform Plaintiffs into agents of the government or cloak them with the authority to speak on its behalf. Plaintiffs remain private speakers just as they were before the law was passed. What then does the law do? Simply put, it requires Plaintiffs to convey a message that the government favors but that Plaintiffs do not. The statute mandates, in unmistakable terms, that private instructors teach (and assume) the validity of the “implicit bias” theory. That is not the government “speak[ing] for itself”; rather, it is the government compelling others to speak in a certain way. Shurtleff v. City of Bos., 596 U.S. 243, 252 (2022). The State  KHATIBI V. HAWKINS 27 of California could have hired its own employees to spread its message of “implicit bias” and the deleterious effects of this purported phenomenon. Matal v. Tam, 582 U.S. 218, 234 (2017) (“The Free Speech Clause does not require government to maintain viewpoint neutrality when its officers and employees speak.”). The State could have enlisted volunteers to do the same. See Shurtleff, 596 U.S. at 270 (Alito, J., concurring in the judgment) (“So long as this responsibility is voluntarily assumed, speech by a private party within the scope of his power to speak for the government constitutes government speech.”). The State could have even created a program in which it was involved “from beginning to end” in proposing edits or suggestions to solicited course material. Matal, 582 U.S. at 237 (citing Johanns v. Livestock Mktg. Ass’n, 544 U.S. 550, 560–61 (2005)); see also Walker v. Texas Div., Sons of Confederate Veterans, Inc., 576 U.S. 200, 213 (2015) (holding that Texas specialty license plates are government speech and considering that “Texas maintains direct control over the messages conveyed on its specialty plates”). All these options might have been considered an exercise in “government speech.” But the State has done none of that. The State has chosen instead to commandeer a vast majority of course providers, who are private, to express a specific viewpoint. That may be the most efficient way for the State to proselytize its message; it may have the added benefit, too, of creating the perception of uniformity on a divisive topic, while imposing a steep social cost on those in the field who dare to dissent. In the end, the statute’s aim appears to be nothing less than ideological conformity enforced through private conscription by the State. 28 KHATIBI V. HAWKINS Our First Amendment stands stubbornly athwart that approach. See W. Va. State Bd. of Educ. v. Barnette, 319 U.S. 624, 642 (1943) (“If there is any fixed star in our constitutional constellation, it is that no official, high or petty, can prescribe what shall be orthodox in politics, nationalism, religion, or other matters of opinion or force citizens to confess by word or act their faith therein.”). As our Supreme Court has said rather bluntly, “the government may not compel a person to speak its own preferred messages,” nor “force an individual to include other ideas with his own speech that he would prefer not to include.” 303 Creative LLC v. Elenis, 600 U.S. 570, 586–87 (2023). “All that offends the First Amendment just the same.” Id. at 587. Yet that appears to be what California is doing here— requiring physicians to treat “implicit bias” as gospel (despite their unwillingness to assent) and, worse still, to then teach it to their students in virtually every continuing education course. Contrary to the panel’s contention (and as Judge VanDyke well explains), the existence of extensive “regulation” in the “medical profession” does not justify State compulsion of a particular viewpoint. See Dissent at 8–11 (VanDyke, J.). If that were so, doctors could be forced to affirm viewpoints they find odious as a condition of maintaining their licenses. Lawyers could find themselves suffering the same fate, too. Both fields (and others) are highly regulated. Indeed, any professional accreditation regime, now open to and supported by a vast network of private providers expressing differing (and perhaps conflicting) viewpoints, would be in jeopardy of being converted into an engine of state-sanctioned groupthink if those providers could be compelled to announce a singular position. That maneuver is not exempt from the First  KHATIBI V. HAWKINS 29 Amendment’s purview. But that is where the panel’s logic leads; for the panel, it is “government speech” all the way down. Nor does private expression become “government speech” simply because that expression is made a condition of a benefit. See, e.g., Matal, 582 U.S. at 239. Being compelled to express a particular viewpoint—on pain of losing professional accreditation—is anything but voluntarily speaking on behalf of the government. See Shurtleff, 596 U.S. at 271 (Alito, J., concurring in the judgment) (citing Matal, 582 U.S. at 234–35) (“Facilitating speech by private persons cannot constitute government speech unless the government assigns a power to speak to those persons or appropriates the products of their expressive activity to express its own message. When the government’s role is limited to applying a standard of assessment to determine a speaker’s eligibility for a benefit, the government is regulating private speech, and ordinary First Amendment principles apply.”). The factors announced in Shurtleff do not help the panel either, as Judge VanDyke correctly concludes. See Shurtleff, 596 U.S. at 252 (“[W]e conduct a holistic inquiry . . . look[ing] to several types of evidence to guide the analysis, including: the history of the expression at issue; the public’s likely perception as to who (the government or a private person) is speaking; and the extent to which the government has actively shaped or controlled the expression.”). But stepping back, the diametrically opposed outcomes that the panel and Judge VanDyke have reached in applying the Shurtleff “test”—and other circuits’ divergence from the panel—might give us pause as to the test’s workability. Id. at 266 (Alito, J., concurring in the judgment) (“[L]ike any factorized analysis, this approach 30 KHATIBI V. HAWKINS cannot provide a principled way of deciding cases.”); see also Dissent at 5, 10–11 (VanDyke, J.). The panel insists that it engaged in Shurtleff’s “holistic inquiry” in concluding that the government speaks—indeed, it invokes the word “holistic” no fewer than nine times in its opinion. But all too often, such “holistic” multi-factor tests serve as mere cover for judges to reach their preferred result. See Lexmark Int’l, Inc. v. Static Control Components, Inc., 572 U.S. 118, 136 (2014) (observing that “open-ended balancing tests[] can yield unpredictable and at times arbitrary results”). Better off, it seems, for constitutional principles to “be anchored in rules, not set adrift in some multifactored ‘balancing test.’” Granfinanciera, S.A. v. Nordberg, 492 U.S. 33, 70 (1989) (Scalia, J., concurring in part and concurring in the judgment). The panel here has gone adrift. If Shurtleff is being applied (as here) to deny First Amendment protection to undeniably private instructors, compelled by the State to teach a doctrine they disbelieve, then something has gone seriously awry and we have lost the plot. The point of the factors is “to determine whether the government intends to speak for itself or to regulate private expression.” Shurtleff, 596 U.S. at 252. If they cease to serve that function reliably, then it is hard to see what function they should serve at all. Here, it is clear the State seeks to “regulate private expression.” I dissent from the denial of rehearing en banc.