Pursuant to 28 U.S.C. § 2342(1) and 47 U.S.C. § 402(a), U.S. West, Inc. petitions for review of a Federal Communication Commission (“FCC”) order restricting the use and disclosure of and access to customer proprietary network information (“CPNI”). See Second Report and Order and Further Notice of Proposed Rulemak-ing: In the Matter of Implementation of the Telecommunications Act of 1996; Telecommunications Carriers’ Use of Consumer Proprietary Network Information and Other Customer Information; Implementation of Non-Accounting Safeguards of Sections 271 and 272 of the Communications Act of 1934, as Amended, 63 Fed. Reg. 20,326 (1998) (“CPNI Order”). Petitioner argues that the regulations adopted by the CPNI Order constitute an arbitrary and capricious interpretation of the controlling provisions of 47 U.S.C. § 222 and are impermissible because they violate the First and Fifth Amendments of the United States Constitution. The regulations require telecommunications companies, in most instances, to obtain affirmative approval from a customer before the company can use that customer’s CPNI for marketing purposes. We vacate the FCC’s CPNI Order, concluding that the FCC failed to adequately consider the constitutional ramifications of the regulations interpreting § 222 and that the regulations violate the First Amendment.
I. Introduction
This case involves classic issues of separation of powers and the courts’ necessary role as guardians of constitutional interests. It is seductive for us to view this as just another case of reviewing agency action. However, this case is a harbinger of difficulties encountered in this age of exploding information, when rights bestowed by the United States Constitution must be guarded as vigilantly as in the days of handbills on public sidewalks. In the name of deference to agency action, important civil liberties, such as the First Amendment’s protection of speech, could easily be overlooked. Policing the boundaries among constitutional guarantees, legislative mandates, and administrative interpretation is at the heart of our responsibility. This case highlights the importance of that role.
II. Background
The dispute in this case involves regulations the FCC promulgated to implement provisions of 47 U.S.C. § 222, which was enacted as part of the Telecommunications Act of 1996. Section 222, entitled “Privacy of customer information,” states generally that “[ejvery telecommunications carrier has a duty to protect the confidentiality of proprietary information of, and relating to ... customers.” 47 U.S.C. § 222(a). To effectuate that duty, § 222 places restrictions on the use, disclosure of, and access to certain customer information. At issue here are the FCC’s regulations clarifying the privacy requirements for CPNI.1 The *1229central provision of § 222 dealing with CPNI is § 222(c)(1), which states:
Except as required by law or with the approval of the customer, a telecommunications carrier that receives or obtains customer proprietary network information by virtue of its provision of a telecommunications service shall only use, disclose, or permit access to individually identifiable customer proprietary network information in its provision of (A) the telecommunication service from which such information is derived, or (B) services necessary to, or used in, the provision of such telecommunications service, including the publishing of directories.
Section 222(d) provides three additional exceptions to the CPNI privacy requirements. Those exceptions allow a telecommunications carrier to use, disclose or permit access to CPNI:
(1) to initiate, render, bill, and collect for telecommunications services,
(2) to protect the rights or property of the carrier, or to protect users of those services and other carriers from fraudulent, abusive, or unlawful use of, or subscription to, such services, or
(3) to provide any inbound telemarketing, referral, or administrative services to the customer for the duration of the call, if such call was initiated by the customer and the customer approves of the use of such information to provide such service.
47 U.S.C. § 222(d). Therefore, the essence of the statutory scheme requires a telecommunications carrier to obtain customer approval when it wishes to use, disclose, or permit access to CPNI in a manner not specifically allowed under § 222.
Section 222 is not the first time the government has placed restrictions on telecommunications carriers’ use or disclosure of CPNI. Prior to the enactment of § 222, the FCC had imposed CPNI requirements on the enhanced service operations of several major telecommunications carriers. See CPNI Order ¶ 7. The FCC imposed these CPNI requirements primarily to prevent large carriers from gaining a competitive advantage in the unregulated enhanced services markets through the use of CPNI, thereby protecting smaller carriers. See id. In contrast, Congress made § 222, which is much broader in scope than previous CPNI requirements, applicable to all carriers, not just the dominant ones. This suggests that Congress enacted § 222 for a substantially different purpose than previous FCC CPNI requirements.
Faced with the new CPNI restrictions, various telecommunications companies and trade associations sought FCC guidance regarding their obligations under § 222. See id. ¶ 6 & n. 25. These requests, along with a petition for a declaratory ruling regarding the interpretation of the term “telecommunication service” under § 222(c)(1), prompted the FCC to commence a rulemaking on May 17, 1996. See id. ¶ 6; In the Matter of Implementation of the Telecommunications Act of 1996: Telecommunication Carriers’ Use of Customer Proprietary Network Information and Other Customer Information, Notice of Proposed Rulemaking, 61 Fed.Reg. 26,-483 (1996) (“CPNI NPRM”). The CPNI NPRM sought comment on, among other things: “(1) the scope of the phrase 'telecommunications service,’ as it is used in section 222(c)(1) ...; (2) the requirements *1230for customer approval; and (3) whether the Commission’s existing CPNI requirements should be amended in light of section 222.” CPNI Order ¶ 6 (citing CPNI NPRM ¶¶ 20-33, 38-42). On February 26, 1998, the FCC released the CPNI Order we now review. The CPNI Order addresses the meaning and scope of § 222 and adopts regulations to implement the statute’s CPNI requirements. See 47 C.F.R. pt. 64, subpt. U (1998).
The regulations adopted by the CPNI Order interpret § 222(c)(1) through a framework known as the “total service approach.” That approach divides the term “telecommunications service” into three service categories: (1) local; (2) interex-change (which includes most long-distance toll service); and (3) commercial mobile radio service (“CMRS”) (which includes mobile or cellular service). See 47 C.F.R. § 64.2006(a). Broadly stated, the regulations permit a telecommunications carrier to use, disclose, or share CPNI for the purpose of marketing products within a category of service to customers, provided the customer already subscribes to that category of service. See id. However, the carrier may not, without customer approval, use, disclose, or permit access to CPNI for the purpose of marketing categories of service to which the customer does not already subscribe. See id. § 64.2005(b).2 For example, petitioner could use CPNI obtained through the provision of local service to market other local service products, but not cellular services. Moreover, if the customer subscribes to both local and long-distance services, petitioner could use the CPNI to market either service and could exchange the CPNI between affiliates that provide such services, but petitioner could still not use the CPNI to market cellular services. In addition, the regulations prevent telecommunications carriers from using, without customer approval, CPNI gained from any of the three categories described above to: (1) market customer premises equipment (“CPE”) or information services (such as call answering, voice mail, or Internet access services); (2) identify or track customers that call competitors; and (3) regain the business of customers who have switched to another carrier. See id. § 64.2005(b)(1)-(3). The regulations also set forth some additional narrow exceptions to the CPNI requirements, other than those stated in § 222(d). See id. § 64.2005(e).
The regulations also describe the means by which a carrier must obtain customer approval. Section 222(c)(1) did not elaborate as to what form that approval should take. The FCC decided to require an “opt-in” approach, in which a carrier must obtain prior express approval from a customer through written, oral, or electronic means before using the customer’s CPNI. See 47 C.F.R. § 64.2007(b). The government acknowledged that the means of approval could have taken numerous other forms, including an “opt-out” approach, in which approval would be inferred from the customer-carrier relationship unless the customer specifically requested that his or her CPNI be restricted.
Petitioner challenges the FCC’s chosen approval process, claiming it violates the First Amendment by restricting its ability to engage in commercial speech with customers. In addition, petitioner argues that the CPNI regulations raise serious Fifth Amendment Takings Clause concerns because CPNI represents valuable property that belongs to the carriers and the regulations greatly diminish its value. The respondents assert that the FCC’s CPNI regulations raise no constitutional concerns, are reasonable, and are entitled to deference under the Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984).
*1231III. Discussion
A. Standard of Review
Under the Administrative Procedure Act, we review a final FCC order to determine whether it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” 5 U.S.C. § 706(2)(A), or “contrary to constitutional right, power, privilege, or immunity,” id. § 706(2)(B). See Long v. Board of Governors of the Fed. Reserve Sys., 117 F.3d 1145, 1151 (10th Cir.1997); City of Albuquerque v. Broumer, 97 F.3d 415, 424 (10th Cir.1996). In addition, when the question before us involves an agency’s interpretation of a statute it administers, we utilize the two-step approach announced in Chevron. See, e.g., Sierra Club v. EPA, 99 F.3d 1551, 1555 (10th Cir.1996). When Congress has spoken to the precise question at issue, we must give effect to the express intent of Congress. See Chevron, 467 U.S. at 842-43, 104 S.Ct. 2778. However, if the statute is silent or ambiguous, we defer to the agency’s interpretation, if it is reasonable. See id. at 843-44, 104 S.Ct. 2778. The agency’s interpretation of the statute need not be the only reasonable or most reasonable interpretation, see id. at 843 n. 11, 104 S.Ct. 2778, but an unconstitutional interpretation is not entitled to Chevron deference.
In addition, deference to an agency interpretation is inappropriate not only when it is conclusively unconstitutional, but also when it raises serious constitutional questions. See Rust v. Sullivan, 500 U.S. 173, 190-91, 111 S.Ct. 1759, 114 L.Ed.2d 233 (1991); Edward J. DeBartolo Corp. v. Florida Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 575-76, 108 S.Ct. 1392, 99 L.Ed.2d 645 (1988); Williams v. Babbitt, 115 F.3d 657, 661-62 (9th Cir.1997), cert. denied sub nom. Kawerak Reindeer Herders Ass’n v. Williams, — U.S. -, 118 S.Ct. 1795, 140 L.Ed.2d 936 (1998); Chamber of Commerce of the United States v. FEC, 69 F.3d 600, 605 (D.C.Cir.1995); Kohler Co. v. Moen Inc., 12 F.3d 632, 634 n. 2 (7th Cir.1993). When faced with a statutory interpretation that ‘would raise serious constitutional problems, the [c]ourt[s] will construe the statute to avoid such problems unless such construction is plainly contrary to the intent of Congress.” DeBartolo Corp., 485 U.S. at 575, 108 S.Ct. 1392. We follow this approach because we assume that Congress legislates with constitutional limitations in mind and will speak clearly when it seeks to test those limitations. See Rust, 500 U.S. at 191, 111 S.Ct. 1759; DeBartolo Corp., 485 U.S. at 575, 108 S.Ct. 1392; Williams, 115 F.3d at 662; International Union, United Auto., Aerospace & Agric. Implement Workers of Am., UAW v. OSHA, 938 F.2d 1310, 1317 (D.C.Cir.1991) (“In effect we require a clear statement by Congress that it intended to test the constitutional waters.”). The Williams court aptly explained the doctrine as it applies to agencies:
[J]ust as we will not infer from an ambiguous statute that Congress meant to encroach on constitutional boundaries, we will not presume from ambiguous language that Congress intended to authorize an agency to do so. At the core of DeBartolo lies the presumption that, if Congress means to push the constitutional envelope, it must do so explicitly.
Williams, 115 F.3d at 662.
Petitioner raises First and Fifth Amendment challenges to the approval procedure adopted by the FCC. The parties agree that Congress did not explicitly set forth the form of customer approval carriers must obtain. Therefore, if we determine that the FCC’s customer approval rule presents a serious or grave constitutional question, we will owe the FCC no deference, even if its CPNI regulations are otherwise reasonable, and will apply the rule of constitutional doubt.
B. Do the CPNI regulations violate the First Amendment?
Petitioner argues that the CPNI regulations interpreting 47 U.S.C. § 222 violate the First Amendment. The First Amendment states, “Congress shall make *1232no law ... abridging the freedom of speech.” U.S. Const, amend. I. Although the text of the First Amendment refers to legislative enactments by Congress, it is actually much broader in scope and encompasses, among other things, regulations promulgated by administrative agencies. See, e.g., Rust, 500 U.S. at 192, 111 S.Ct. 1759 (subjecting Department of Health and Human Services regulations limiting the ability of Title X fund recipients to engage in abortion-related activities to review under the First Amendment).
1. Do the CPNI regulations restrict speech?
As a threshold requirement for the application of the First Amendment, the government action must abridge or restrict protected speech. The government argues that the FCC’s CPNI regulations do not violate or even infringe upon petitioner’s First Amendment rights because they only prohibit it from using CPNI to target customers and do not prevent petitioner from communicating with its customers or limit anything that it might say to them. This view is fundamentally flawed. Effective speech has two components: a speaker and an audience. A restriction on either of these components is a restriction on speech. Cf. Virginia State Bd. of Pharmacy v. Virginia Citizens Consumer Council, Inc., 425 U.S. 748, 756-57, 96 S.Ct. 1817, 48 L.Ed.2d 346 (1976) (noting that the First Amendment protects the communication, whether the speech restriction applies to its source or impinges upon the audience’s reciprocal right to receive the communication); Martin v. City of Struthers, 319 U.S. 141, 143, 63 S.Ct. 862, 87 L.Ed. 1313 (1943) (noting the First Amendment “embraces the right to distribute literature and necessarily protects the right to receive it”). In other words, a restriction on speech tailored to a particular audience, “targeted speech,” cannot be cured simply by the fact that a speaker can speak to a larger indiscriminate audience, “broadcast speech.”
Perhaps the Supreme Court case of Florida Bar v. Went For It, Inc., 515 U.S. 618, 115 S.Ct. 2371, 132 L.Ed.2d 541 (1995), best illustrates this. In Went For It, a lawyer referral service and an individual lawyer challenged a Florida Bar rule that prohibited attorneys from using direct mail advertisements to solicit wrongful death and personal injury clients within thirty days of the accident or disaster causing death or injury. See 515 U.S. at 620-21, 115 S.Ct. 2371. Despite the fact that the attorney could indiscriminately mail solicitations for his services, the court found that the targeted speech constituted commercial speech and that the restriction on the targeted speech implicated the First Amendment. See id. at 623, 115 S.Ct. 23713; see also Ficker v. Curran, 119 F.3d 1150, 1153-56 (4th Cir.1997) (applying First Amendment analysis to direct mail solicitations by attorneys to criminal and traffic defendants); Revo v. Disciplinary Bd. of the Sup.Ct. for the State of N.M., 106 F.3d 929, 932-33 (10th Cir.) (determining that lawyer’s direct mail advertising to personal injury victims and family members of wrongful death victims constituted protected commercial speech), cert. denied, 521 U.S. 1121, 117 S.Ct. 2515, 138 L.Ed.2d 1017 (1997). Therefore, the existence of alternative channels of communication, such as broadcast speech, does not eliminate the fact that the CPNI regulations restrict speech.
2. What kind of speech is restricted?
Because petitioner’s targeted speech to its customers is for the purpose of soliciting those customers to purchase more or different telecommunications services, it “does no more than propose a commercial transaction,” Virginia State Bd. of Pharmacy, 425 U.S. at 760, 96 S.Ct. 1817 (quoting Pittsburgh Press Co. v. Human Relations Comm’n, 413 U.S. 376, 385, *123393 S.Ct. 2558, 37 L.Ed.2d 669 (1973)). Consequently, the targeted speech in this case fits soundly within the definition of commercial speech. See id.; Bolger v. Youngs Drug Prods. Corp., 463 U.S. 60, 66, 103 S.Ct. 2875, 77 L.Ed.2d 469 (1983); Cardtoons, L.C. v. Major League Baseball Players Ass’n, 95 F.3d 959, 970 (10th Cir.1996); see also, e.g., Bad Frog Brewery, Inc. v. New York State Liquor Auth., 134 F.3d 87, 97 (2d Cir.1998) (“The ‘core notion’ of commercial speech includes speech which does no more than propose a commercial transaction.” (internal quotation marks and citation omitted)). It is well established that nonmisleading commercial speech regarding a lawful activity is a form of protected speech under the First Amendment, although it is generally afforded less protection than noncommercial speech. See, e.g., Went For It, 515 U.S. at 623, 115 S.Ct. 2371; Central Hudson Gas & Elec. Corp. v. Public Serv. Comm’n of N.Y., 447 U.S. 557, 562-63, 100 S.Ct. 2343, 65 L.Ed.2d 341 (1980). The parties do not dispute that the commercial speech based on CPNI is truthful and nonmisleading. Therefore, the CPNI regulations implicate the First Amendment by restricting protected commercial speech.4
3. Central Hudson analysis
We analyze whether a government restriction on commercial speech violates the First Amendment under the four-part framework set forth in Central Hudson. First, we must conduct a threshold inquiry regarding whether the commercial speech concerns lawful activity and is not misleading. See Central Hudson, 447 U.S. at 566, 100 S.Ct. 2343. If these requirements are not met, the government may freely regulate the speech. See Went For It, 515 U.S. at 623-24, 115 S.Ct. 2371; Revo, 106 F.3d at 932. If this threshold requirement is met, the government may restrict the speech only if it proves: “(1) it has a substantial state interest in regulating the speech, (2) the regulation directly and materially advances that interest, and (3) the regulation is no more extensive than necessary to serve the interest.” Revo, 106 F.3d at 932 (citing Central Hudson, 447 U.S. at 564-65, 100 S.Ct. 2343).5 *1234As noted above, no one disputes that the commercial speech based on CPNI is truthful and nonmisleading. We therefore proceed directly to whether the government has satisfied its burden under the remaining three prongs of the Central Hudson test.
a. Does the government have a substantial state interest in regulating speech involving CPNI?
The respondents argue that the FCC’s CPNI regulations advance two substantial state interests: protecting customer privacy and promoting competition. While, in the abstract, these may constitute legitimate and substantial interests, we have concerns about the proffered justifications in the context of this case.
Privacy considerations of some sort clearly drove the enactment of § 222. The concept of privacy, though, is multi-facet-ed. Indeed, one can apply the moniker of a privacy interest to several understandings of privacy, such as the right to have sufficient moral freedom to exercise full individual autonomy, the right of an individual to define who he or she is by controlling access to information about him or herself, and the right of an individual to solitude, secrecy, and anonymity.6 See Fred H. Cate, Privacy in the Information Age 19-22 (1997); Joseph I. Rosenbaum, Privacy on the Internet: Whose Information Is It Anyway?, 38 Jurimetrics J. 565, 566-67 (1998). The breadth of the concept of privacy requires us to pay particular attention to attempts by the government to assert privacy as a substantial state interest.
When faced with a constitutional challenge, the government bears the responsibility of building a record adequate to clearly articulate and justify the state interest. “[T]he Central Hudson standard does not permit us to supplant the precise interests put forward by the State with other suppositions.” Edenfield v. Fane, 507 U.S. 761, 768, 113 S.Ct. 1792, 123 L.Ed.2d 543 (1993). Although we agree that privacy may rise to the level of a substantial state interest, see, e.g., Went For It, 515 U.S. at 625, 115 S.Ct. 2371 (“Our precedents leave no room for doubt that ‘the protection of potential clients’ privacy is a substantial state interest’ ” (quoting Edenfield, 507 U.S. at 769, 113 S.Ct. 1792)), the government cannot satisfy the second prong of the Central Hudson *1235test by merely asserting a broad interest in privacy. It must specify the particular notion of privacy and interest served. Moreover, privacy is not an absolute good because it imposes real costs on society.7 Therefore, the specific privacy interest must be substantial, demonstrating that the state has considered the proper balancing of the benefits and harms of privacy. In sum, privacy may only constitute a substantial state interest if the government specifically articulates and properly justifies it.
In the context of a speech restriction imposed to protect privacy by keeping certain information confidential, the government must show that the dissemination of the information desired to be kept private would inflict specific and significant harm on individuals, such as undue embarrassment or ridicule, intimidation or harassment, or misappropriation of sensitive personal information for the purposes of assuming another’s identity. Although we may feel uncomfortable knowing that our personal information is circulating in the world, we live in an open society where information may usually pass freely. A general level of discomfort from knowing that people can readily access information about us does not necessarily rise to the level of a substantial state interest under Central Hudson for it is not based on an identified harm.
Neither Congress nor the FCC explicitly stated what “privacy” harm § 222 seeks to protect against. The CPNI Order notes that “CPNI includes information that is extremely personal to customers ... such as to whom, where, and when a customer places a call, as well as the types of service offerings to which the customer subscribes,” CPNI Order at ¶ 2, and it summarily finds “call destinations and other details about a call ... may be equally or more sensitive [than the content of the calls],” id. at ¶ 94. The government never states it directly, but we infer from this thin justification that disclosure of CPNI information could prove embarrassing to some and that the government seeks to combat this potential harm.
We have some doubts about whether this interest, as presented, rises to the level of “substantial.” We would prefer to see a more empirical explanation and justification for the government’s asserted interest. Cf. Went For It, 515 U.S. at 630, 115 S.Ct. 2371 (describing the record provided by the Bar cataloguing citizen outrage at being solicited just after injury or family tragedy). In addition, the authority relied upon by the government, Edenfield v. Fane, recognizes a state’s interest in protecting against unwanted intrusions caused by solicitations, see 507 U.S. at 769, 113 S.Ct. 1792; see also Went For It, 515 U.S. at 625, 115 S.Ct. 2371, but it says nothing about the disclosure of allegedly sensitive information. On the other hand, we recognize the government may have a legitimate interest in helping protect certain information. Cf. Lanphere & Urbaniak v. Colorado, 21 F.3d 1508, 1514 (10th Cir.1994) (finding a substantial state interest in the need to protect the privacy of those charged with traffic offenses and DUI against dissemination of charging information for commercial purposes). Therefore, notwithstanding our reserva*1236tions, we assume for the sake of this appeal that the government has asserted a substantial state interest in protecting people from the disclosure of sensitive and potentially embarrassing personal information.8
We harbor different reservations about the government’s asserted interest in competition. While we afford agencies broad deference in interpreting a statute they are charged to administer, they must obey the dictates of Congress and administer the statute true to Congress’ intent. See Ernst & Ernst v. Hochfelder, 425 U.S. 185, 213-14, 96 S.Ct. 1375, 47 L.Ed.2d 668 (1976). We are not satisfied that the interest in promoting competition was a significant consideration in the enactment of § 222.
While the broad purpose of the Telecommunications Act of 1996 is to foster increased competition in the telecommunications industry,9 the language of § 222 reveals no such concern.10 Rather, the specific and dominant purpose of § 222 is the protection of customer privacy. Indeed, the FCC and members of Congress characterize § 222 as “striving] to balance both the competitive and consumer privacy interests with respect to CPNI,” Joint Statement of Managers, S. Conf. Rep. No. 104-230, at 205 (1996) (emphasis added), which suggests that § 222’s purpose in fostering privacy may even run counter to the broad pro-competition purpose of the Telecommunications Act. In any event, three other considerations persuade us that Congress did not intend for competition to be a significant purpose of § 222. First, and most important, the plain language of the section deals almost exclusively with privacy. Section 222 is entitled “Privacy of customer information” and is replete with references to privacy and confidentiality of customer information. In contrast, § 222 contains no explicit mention of competition. Although § 222(c)(3) and § 222(e) impose nondiscrimination requirements with respect to dis*1237closure of aggregate customer and subscriber list information which could be construed as pro-competition measures, we find that these do not sufficiently indicate that increasing competition was a purpose of § 222. Moreover, the provisions of § 222 relating to CPNI which the challenged regulations interpret contain no reference to nondiscrimination requirements and reflect solely a concern for customer privacy. See 47 U.S.C. § 222(c)(1)-(2), (d). Second, § 222 differs from previous CPNI restrictions designed to foster competition because it applies to all telecommunications carriers, not just the dominant ones. This indicates a different purpose for the new restriction. Finally, § 222 contains measures that will allow full use, disclosure, and access to CPNI if customer approval is obtained. Assuming that a carrier is able to obtain a high rate of customer approval, the alleged competitive effect of § 222’s CPNI restrictions is minimal and can perhaps even be nullified. Consequently, we find that Congress’ primary purpose in enacting § 222 was concern for customer privacy, not the broader purpose of increasing competition.
Even though we conclude that competition did not constitute the primary purpose of the section, we recognize that Congress may not have completely ignored competition in drafting § 222. While we believe that the asserted interest in increasing competition would not suffice, by itself, to justify the FCC’s rule, we will, in this case, consider it in concert with the government’s interest in protecting consumer privacy.
b. Does the Regulation Directly and Materially Advance the State’s Interests?
Under the next prong of Central Hudson, the government must “demonstrate that the harms it recites are real and that its restriction will in fact alleviate them to a material degree.” Edenfield v. Fane, 507 U.S. 761, 771, 113 S.Ct. 1792, 123 L.Ed.2d 543 (1993); accord Rubin v. Coors Brewing Co., 514 U.S. 476, 487, 115 S.Ct. 1585, 131 L.Ed.2d 532 (1995). “This burden is not satisfied by mere speculation or conjecture.” Edenfield, 507 U.S. at 770, 113 S.Ct. 1792. On the record before us, the government fails to meet its burden.
The government presents no evidence showing the harm to either privacy or competition is real. Instead, the government relies on speculation that harm to privacy and competition for new services will result if carriers use CPNI. In Eden-field, the Supreme Court struck down a Florida ban on CPA in-person solicitation because the state had presented no evidence — anecdotal or empirical — that such solicitation created the dangers of “fraud, overreaching, or compromised independence” that the state sought to combat. See 507 U.S. at 771, 113 S.Ct. 1792; cf. Florida Bar v. Went For It, Inc., 515 U.S. 618, 626-27, 115 S.Ct. 2371, 132 L.Ed.2d 541 (1995) (upholding restriction on solicitation of accident victims within thirty days of accident, based on two-year study and written report analyzing statistically and anecdotally the impacts of such solicitation). The FCC faces the same problem here. While protecting against disclosure of sensitive and potentially embarrassing personal information may be important in the abstract, we have no indication of how it may occur in reality with respect to CPNI. Indeed, we do not even have indication that the disclosure might actually occur. The government presents no evidence regarding how and to whom carriers would disclose CPNI. By its own admission, the government is not concerned about the disclosure of CPNI within a firm. See CPNI Order at ¶ 55, n. 203 (“[W]e agree ... that sharing of CPNI within one integrated firm does not raise significant privacy concerns because customers would not be concerned with having their CPNI disclosed within a firm in order to receive increased competitive offerings.”). Yet the government has not explained how or why a carrier would disclose CPNI to outside parties, especially when the government claims CPNI is information that would give one firm a competitive advantage over another. This *1238leaves us unsure exactly who would potentially receive the sensitive information.
Similarly, the FCC can theorize that allowing existing carriers to market new services with CPNI will impede competition for those services, but it provides no analysis of how or if this might actually occur. Beyond its own speculation, the best the government can offer is that “[t]he vigor of U.S. West’s protests against the rules ... indicates that U.S. West also believes that this restriction will be effective in promoting Congress’s competitive interest.” Appellees Br. at 30. This is simply additional conjecture, and it is inadequate to justify restrictions under the First Amendment. See Edenfield, 507 U.S. at 770-71, 113 S.Ct. 1792.
c. Are the CPNI regulations narrowly tailored?
Even assuming, arguendo, that the state interests in privacy and competition are substantial and that the regulations directly and materially advance those interests, we do not find, on this record, the FCC rules regarding customer approval properly tailored. The CPNI regulations must be “no more extensive than necessary to serve [the stated] interest[s].” Rubin, 514 U.S. at 486, 115 S.Ct. 1585. In order for a regulation to satisfy this final Central Hudson prong, there must be a fit between the legislature’s means and its desired objective—“a fit that is not necessarily perfect, but reasonable; that represents not necessarily the single best disposition but one whose scope is in proportion to the interest served.” Board of Trustees of the State Univ. of N.Y. v. Fox, 492 U.S. 469, 480, 109 S.Ct. 3028, 106 L.Ed.2d 388 (1989) (internal quotation marks omitted). While clearly the government need not employ the least restrictive means to accomplish its goal, it must utilize a means that is “narrowly tailored” to its desired objective. Id.; Florida Bar v. Went For It, Inc., 515 U.S. 618, 632, 115 S.Ct. 2371, 132 L.Ed.2d 541 (1995). Narrow tailoring means that the government’s speech restriction must signify a “carefu[l] calculation of] the costs and benefits associated with the burden on speech imposed by its prohibition.” Cincinnati v. Discovery Network, Inc., 507 U.S. 410, 417, 113 S.Ct. 1505, 123 L.Ed.2d 99 (1993) (internal quotation marks omitted). “The availability of less burdensome alternatives to reach the stated goal signals that the fit between the legislature’s ends and the means chosen to accomplish those ends may be too imprecise to withstand First Amendment scrutiny.” 44 Liquormart, Inc. v. Rhode Island, 517 U.S. 484, 529, 116 S.Ct. 1495, 134 L.Ed.2d 711 (1996) (O’Connor, J., concurring); see also, e.g., Romer v. Evans, 517 U.S. at 632, 116 S.Ct. 1620; Rubin, 514 U.S. at 490-91, 115 S.Ct. 1585; Discovery Network, 507 U.S. at 417 n. 13, 113 S.Ct. 1505. This is particularly true when such alternatives are obvious and restrict substantially less speech.11 See Fox, 492 U.S. at 479, 109 S.Ct. 3028 (“[A]lmost all of the restrictions disallowed under Central Hudson ’s fourth prong have been substantially excessive, disregarding ‘far less restrictive and more precise means.’ ” (quoting Shapero v. Kentucky Bar Ass’n, 486 U.S. 466, 476, 108 S.Ct. 1916, 100 L.Ed.2d 475 (1988))).
It is difficult, if not impossible, for us to conduct a full and proper narrow tailoring analysis, given the deficiencies that we have already encountered with respect to the previous portions of the Central Hudson test. Nevertheless, on this record, the FCC’s failure to adequately consider an obvious and substantially less restrictive alternative, an opt-out strategy, indicates that it did not narrowly *1239tailor tbe CPNI regulations regarding customer approval. The respondents argue that the record contains adequate support that the CPNI regulations are narrowly tailored because a study conducted by petitioner U.S. West shows that a majority of individuals, when affirmatively asked for approval to use CPNI, refused to grant it. The U.S. West study shows that 33% of those called refused to grant approval to use their CPNI, 28% granted such approval, and 39% either hung up or asked not to be called again. See CPNI Order ¶ 99 n. 380. Additionally, U.S. West secured a 72% affirmative response rate from customers whom it solicited after they initiated contact with the company for some other reason.12 See id. ¶ 99 n. 378. This study does not provide sufficient evidence that customers do not want carriers to use their CPNI. The results may simply reflect that a substantial number of individuals are ambivalent or disinterested in the privacy of their CPNI or that consumers are averse to marketing generally. The FCC stated that the study supported “an equally plausible interpretation ... that many customers value the privacy of their personal information and do not want it shared for purposes beyond the existing service relationship.” CPNI Order ¶ 100. We are not convinced that the study supports the FCC’s interpretation, and the FCC provides no additional evidence to bolster its argument.
Even assuming that telecommunications customers value the privacy of CPNI, the FCC record does not adequately show that an opt-out strategy would not sufficiently protect customer privacy. The respondents merely speculate that there are a substantial number of individuals who feel strongly about their privacy, yet would not bother to opt-out if given notice and the opportunity to do so. Such speculation hardly reflects the careful calculation of costs and benefits that our commercial speech jurisprudence requires.
Finally, respondents assert that under FCC v. National Citizens Comm. for Broad., 436 U.S. 775, 98 S.Ct. 2096, 56 L.Ed.2d 697 (1978), the FCC can rely upon its common sense judgment based on experience, notwithstanding the inclusiveness of the rulemaking record. We refuse to extend the rule announced in National Citizens in the manner respondents suggest. National Citizens involved agency conclusions regarding “elusive concepts, not easily defined let alone measured without making qualitative judgments,” id. at 796-97, 98 S.Ct. 2096 (internal quotation marks omitted), and information that was difficult to compile. We see no such problems in this case. Furthermore, in National Citizens, the FCC’s common sense judgment only supported a finding that it “acted rationally” in promulgating a rule. Id. at 796, 98 S.Ct. 2096. The burden under the fourth prong of Central Hudson is significantly higher. The FCC must not only demonstrate that it acted rationally, but that it narrowly tailored its regulations to meet its stated goals.
In sum, even assuming that respondents met the prior two prongs of Central Hudson, we conclude that based on the record before us, the agency has failed to satisfy its burden of showing that the customer approval regulations restrict no more speech than necessary to serve the asserted state interests.13 Consequently, we find that the CPNI regulations interpreting the customer approval requirement of 47 U.S.C. § 222(c) violate the First Amendment.14
*1240IV. Conclusion
The FCC failed to adequately consider the constitutional implications of its CPNI regulations. Even if we accept the government’s proffered interests and assume those interests are substantial, the FCC still insufficiently justified its choice to adopt an opt-in regime. Consequently, its CPNI regulations must fall under the First Amendment. At the very least, the foregoing analysis shows that the CPNI regulations clearly raise a serious constitutional question, invoking the rule of constitutional doubt. Accordingly, we VACATE the FCC’s CPNI Order and the regulations adopted therein.15
. The statute recognizes three types of customer information: (1) CPNI; (2) aggregate customer information; and (3) subscriber list information. The statute defines CPNI as:
(A) information that relates to the quantity, technical configuration, type, destination, and amount of use of a telecommunications service subscribed to by any customer of a telecommunications carrier, and that is made available to the carrier by the customer solely by virtue of the carrier-customer relationship; and
(B) information contained in the bills pertaining to telephone exchange service or telephone tool service received by a customer of a carrier;
except that such term does not include subscriber list information.
*122947 U.S.C. § 222(0(1 )(A)-(B). Given the sensitive nature of some CPNI, such as when, where, and to whom a customer places calls, Congress afforded CPNI the highest level of privacy protection under § 222. By way of comparison, aggregate customer information is "collective data that relates to a group or category of services or customers, from which individual customer identities and characteristics have been removed,” id. § 222(f)(2), and subscriber list information consists of the type of information normally published in telephone directories, such as names, numbers, addresses and primary advertising classifications, see id. § 222(0(3). Congress afforded these other types of customer information substantially less privacy protection under § 222. See id. §§ 222(c)(3), (e).
. The regulations treat affiliated entities of a carrier as separate for the purposes of use or disclosure. Thus, the regulations permit unapproved disclosure of CPNI between affiliated entities of a telecommunications carrier only when the carrier provides different categories of service and the customer subscribes to more than one category of service. See id. §§ 64.2005(a)(l)-(2).
. The court did, however, consider this fact in determining whether the speech restriction was narrowly tailored. See id. at 633-34.
. Petitioner argues that because the CPNI regulations also burden its internal business communications (e.g., communications between its affiliates, divisions, and employees), we should subject the regulations to the more stringent level of First Amendment scrutiny applied to restrictions on noncommercial speech. Without deciding whether the incidental burden on internal business communications necessarily implicates the First Amendment or whether petitioner has standing to assert such an argument, we find that, in this case, the intra-carrier speech is properly categorized as commercial speech and consequently its existence does not impact our analysis.
Petitioner asserts that the intra-carrier speech does not directly propose a commercial transaction to customers and therefore falls outside the definition of commercial speech. We disagree. Although speech that merely proposes a commercial transaction is at the "core” of commercial speech, it does not constitute the universe of commercial speech. Indeed, the Supreme Court has defined commercial speech in broader terms as "expression related solely to the economic interests of the speaker and its audience.” Central Hudson, 447 U.S. at 561, 100 S.Ct. 2343. It is, admittedly, unclear to what extent Central Hudson broadened the definition of commercial speech. As another circuit recently stated, “the Court has not offered any nuanced distinctions between the two standards [, i.e., the Virginia Pharmacy and Central Hudson definitions], and the Court noted in Discovery Network that it had not utilized the broader test in its recent commercial speech cases.” Commodity Trend Serv., Inc. v. Commodity Futures Trading Comm’n, 149 F.3d 679, 685 (7th Cir.1998) (citing City of Cincinnati v. Discovery Network, Inc., 507 U.S. 410, 422, 113 S.Ct. 1505, 123 L.Ed.2d 99 (1993)). Consequently, we are hesitant to broadly expand the definition of commercial speech. However, in this case, when the sole purpose of the intra-carrier speech based on CPNI is to facilitate the marketing of telecommunications services to individual customers, we find the speech integral to and inseparable from the ultimate commercial solicitation. Therefore, the speech is properly categorized as commercial speech.
. In 44 Liquormart, Inc. v. Rhode Island, 517 U.S. 484, 116 S.Ct. 1495, 134 L.Ed.2d 711 (1996), the Supreme Court established a slight modification to the Central Hudson framework by giving force to a footnote contained in Central Hudson. Justice Stevens, *1234writing for a four Justice plurality, stated that when a regulation constitutes a blanket prohibition against truthful, nonmisleading speech about a lawful product and the ban serves an interest unrelated to consumer protection, it will be subject to a heightened form of First Amendment scrutiny akin to strict scrutiny. See id. at 504 (opinion of Stevens, J.) (citing Central Hudson, 447 U.S. at 566 n. 9, 100 S.Ct. 2343). Under such circumstances, we must review the regulation under Central Hudson with "special care, mindful that speech prohibitions of this type rarely survive constitutional review." Id. (opinion of Stevens, J.) (internal quotation marks and citation omitted). Although only four Justices subscribed to this view, given Justice Thomas’ concurrence in which he stated that he would abandon Central Hudson altogether and apply traditional strict scrutiny under similar circumstances, see id. at 518, 116 S.Ct. 1495 (Thomas, J., concurring), it is the narrowest majority holding, and we are bound by it.
In this case, however, the regulation at issue does not constitute a blanket prohibition of speech. Indeed, the telecommunications carriers may utilize a multitude of communication channels to say whatever they want to their customers. They simply cannot use CPNI to target customers for marketing efforts. Thus, the CPNI regulations are not subject to heightened scrutiny under 44 Li-quormart.
. We emphasize that the privacy interest in this case is distinct and different from the more limited notion of a constitutional right to privacy which is addressed in cases such as Griswold v. Connecticut, 381 U.S. 479, 484-86, 85 S.Ct. 1678, 14 L.Ed.2d 510 (1965), and Roe v. Wade, 410 U.S. 113, 152-56, 93 S.Ct. 705, 35 L.Ed.2d 147 (1973) (stating that the constitutional right to privacy covers only personal rights deemed "fundamental” or "implicit in the concept of ordered liberty” (internal quotation marks and citations omitted)). Here, the question is solely whether privacy can constitute a substantial state interest under Central Hudson, not whether the FCC regulations impinge upon an individual’s right to privacy under the Constitution.
. Professor Cate lists a number of costs privacy imposes. For example, privacy "facilitates the dissemination of false information,” by making it more difficult for individuals and institutions to discover falsities. Cate, supra, at 28. Privacy also "protects the withholding of relevant true information,” such as when an employee fails to disclose a medical condition that would affect his or her job performance. Id. In addition, privacy interferes with the collection, organization, and storage of information which can assist businesses in making rapid, informed decisions and efficiently marketing their products or services. In this sense, privacy may lead to reduced productivity and higher prices for those products or services. See id. at 28-29. Privacy may even threaten physical safety by interfering with the public’s ability to access information needed to protect themselves, such as whether an individual has a history of child abuse or molestation, sexual offenses, or communicable diseases. See id. at 29. Finally, privacy impedes upon individual voyeuristic curiosity which "opens people’s eyes to opportunities and dangers.” Id. at 29-30.
.In its brief and at oral argument, the FCC intimated that consumer privacy concerns might also encompass an interest in preventing the customer intrusion that accompanies broad use of CPNI for telemarketing purposes. However, this particular privacy justification is lacking from the FCC record. In fact, the only reference to marketing intrusion into customer privacy comes in paragraph 100 of the CPNI Order in response to U.S. West’s attempts to explain why it had such difficulty obtaining authorization to use CPNI in a telemarketing and direct-mail study it conducted. In paragraph 100, the FCC stated:
[E]ven if U S WEST is correct, and customers do not grant approval simply because they do not want to be marketed to, this finding would not support permitting notice and opt-out. Indeed, it would suggest, as MCI observes, that contrary to U S WEST’s claim, customers do not want to hear about "expanding service offerings," and in particular do not want their CPNI used towards that end.
CPNI Order ¶ 100. Such a terse statement, made only in the limited context of refuting U.S. West's expansive reading of its prior market study, provides insufficient evidence that the FCC sought to promote customer privacy by limiting intrusion into customer households through telemarketing made possible by CPNI sharing.
. The preamble to the Act states: "An act to promote competition and reduce regulation in order to secure lower prices and higher quality services for American telecommunications consumers and encourage the rapid deployment of new telecommunications technologies.” Pub.L. No. 104-104, 110 Stat. 56, 56 (1996).
. While the broad purposes of an Act frequently provide useful insight into the purposes served by a narrow provision of the Act, blind adherence to broad purposes can obfuscate Congress' true intent regarding a particular provision, particularly when that provision has an unambiguous, specific, and dominant purpose. See Board of Governors of the Fed. Reserve Sys. v. Dimension Fin. Corp., 474 U.S. 361, 373-74, 106 S.Ct. 681, 88 L.Ed.2d 691 (1986) (“Application of 'broad purposes' of legislation at the expense of specific provisions ignores the complexity of the problems Congress is called upon to address and the dynamics of legislative action.... Invocation of the 'plain purpose' of legislation at the expense of the terms of the statute itself takes no account of the processes of compromise and, in the end, prevents the effectuation of congressional intent.”).
. While this pronouncement, in effect, imposes a burden on the government to consider certain less restrictive means—those that are obvious and restrict substantially less speech—it does not amount to a least restrictive means test. We do not require the government to consider every conceivable means that may restrict less speech and strike down regulations when any less restrictive means would sufficiently serve the state interest. We merely recognize the reality that the existence of an obvious and substantially less restrictive means for advancing the desired government objective indicates a lack of narrow tailoring.
. U.S. West also solicited approval from customers by mail. Only six to eleven percent of residential customers and only five to nine percent of business customers responded to the direct mail trial. See CPNI Order ¶ 99 n. 378.
. We reiterate that even if the opt-in approach is narrowly tailored with respect to protecting competition, the interest advanced in protecting competition here is insufficient by itself to justify the CPNI regulations under the Central Hudson test. See supra Part III. B.3.a.
.Because we vacate the CPNI restrictions on First Amendment grounds, we need not address whether they have effected a "taking” under the Fifth Amendment or whether they are otherwise arbitrary and capricious.
. The dissent accuses us of “advocating" an opt-out approach. We do not “advocate” any specific approach. We merely find fault in the FCC’s inadequate consideration of the approval mechanism alternatives in light of the First Amendment.