Symposium - 2018 - #MeToo

Targeting Repeat Offender NDAs

Ian Ayres*

Abstract. While nondisclosure/non-disparagement agreements (NDAs) can beneficially protect privacy and facilitate settlement of sexual misconduct claims, these agreements have come under attack—especially since the rise of the #MeToo movement—because NDAs can also facilitate repeat offending. While some academics and policy makers have proposed making NDAs unenforceable, this Essay searches for intermediate legal interventions which preserve some of the privacy benefits while targeting the problem of repeat offending. I propose that NDAs should only be enforceable if they meet particular disclosure, condition, and escrow requirements. Specifically, the Essay proposes that NDAs only be enforceable (1) if they explicitly disclose the rights which the survivor retains, notwithstanding the NDA, to report the perpetrator’s behavior to the Equal Employment Opportunity Commission (EEOC) and other investigative authorities; (2) if they include a condition that any misrepresentation by the perpetrator about the survivor and perpetrator’s past interactions will constitute a material breach giving the survivor the option to cancel the NDA and speak about those past interactions; and (3) if the underlying survivor allegations are deposited in an information escrow that would be released for investigation by the EEOC (or the Department of Education with regard to Title IX violations) and other investigative authorities if another complaint is received against the same perpetrator.


Recent months have seen media reports of serial settlement of sexual misconduct claims concerning several powerful men. The New York Times reported on six settlement agreements, totaling about $45 million, related to claims against former Fox News host Bill O’Reilly. 1 See Emily Steel & Michael S. Schmidt, Bill O’Reilly Thrives at Fox News, Even as Harassment Settlements Add Up, N.Y. Times (Apr. 1, 2017), (describing five settlements between Bill O’Reilly or Fox News and women for alleged sexual misconduct or other inappropriate behavior); Emily Steel & Michael S. Schmidt, Bill O’Reilly Settled New Harassment Claim, Then Fox Renewed His Contract, N.Y. Times (Oct. 21, 2017), (detailing a sixth settlement). The New Yorker detailed a series of settlements spanning two decades involving claims against film producer Harvey Weinstein. 2 See Ronan Farrow, Harvey Weinstein’s Secret Settlements, New Yorker (Nov. 21, 2017), Additionally, several women stated in television interviews that they entered into settlements and contracts that prevented them from discussing their claims of extramarital relationships with President Donald Trump. 3 Cf. Jim Rutenberg, Ex-Playboy Model, Freed from Contract, Can Discuss Alleged Trump Affair, N.Y. Times (Apr. 18, 2018), (describing the agreement that prevented Karen McDougal from discussing an alleged relationship with President Trump); Christie Smythe, Stormy Daniels Offers to Pay Back Trump Settlement to End Silence, Bloomberg (Mar. 12, 2018),

The inclusion in these settlement agreements of nondisclosure promises from survivors has raised concerns that these agreements can be used to shield repeat offenders from investigation—allowing them to offend again and again. But nondisclosure agreements (NDAs) can have legitimate uses. Some survivors want privacy. Survivors can reasonably fear that being known as a person who makes sexual misconduct allegations will reduce their future employment prospects or lead to being accused or suspected of lying or a variety of other negative consequences. NDAs may also help protect those who are falsely accused or have a valid legal defense from the negative reputational consequences of having been accused and having paid to settle an accusation of sexual misconduct.

NDAs may also be jointly valued by the accuser and the accused for a less clearly legitimate reason. When unlawful sexual misconduct did in fact occur, the promise of nondisclosure might be valued by the survivor solely because it secures for the survivor a larger settlement payment, and valued by the perpetrator as a shield from public scrutiny for the prior wrongdoing. While NDAs might be mutually beneficial for the contractors, they might make it easier for the perpetrator to reoffend. A central concern with enforcement is that NDAs do not adequately manifest the assent of potential future victims of the perpetrator.

The prospect that NDAs might shield repeat offenders from investigation has—especially since the revelation of high profile serial offenders protecting themselves from public scrutiny by using a series of NDAs—led some policy makers to propose reforms to limit the enforceability of such provisions. California, for example, has introduced a bill to ban NDAs in settlements for sexual assault and harassment. 4 Press Release, Senator Leyva Introduces STAND (Stand Together Against Non-Disclosures) Act (Jan. 3, 2018), Lawmakers in New York, Pennsylvania, and New Jersey have also proposed similar bans. 5 Daniel Wiessner, Sexual Misconduct Outcry Makes U.S. Lawyers Rethink Confidentiality, Reuters (Dec. 19, 2017, 4:08 AM),

NDAs seem to present us with an all-or-nothing choice. We can decide either that the legitimate (survivor privacy or false accusation) rationales predominate and continue to enforce these provisions, or that the shielded repeat offender rationale predominates and refuse to enforce NDAs across the board.

This Essay instead undertakes the difficult task of proposing an intermediate reform that seeks to allow many of the legitimate NDA benefits but curtails some shielding of serial offenders. The task is daunting because it is difficult to target NDAs that are being used by repeat offenders. The NDA provisions of course don’t reveal this fact. Indeed, from the serial offender’s perspective, an important goal of the NDA is to protect the offender from having to reveal the serial nature of their activity. From a policy perspective, we might want to particularly distinguish one-off NDA settlements from those where the offender is a potential reoffender. But since all offenders are potential reoffenders, to make such a distinction would require identifying the even more elusive and contestable category of NDAs entered into by the falsely accused. Policy makers seem to have little ability to make such fine-grained distinctions.

The difficulty of distinguishing and targeting these different types of NDAs can be seen in the recent tax reform act. The act, which became law in December of 2017 and remains one of the most notable accomplishments of the Trump administration, contains a provision that eliminates the business deduction for any settlement of sexual misconduct claims that includes an NDA. 6 The amended statute prevents the deduction of “(1) any settlement or payment related to sexual harassment or sexual abuse . . . subject to a nondisclosure agreement, or (2) attorney’s fees related to such a settlement or payment” as business expenses. Act of Dec. 22, 2017, Pub. L. No. 115-97, § 13307, 131 Stat. 2054, 2129 (to be codified at I.R.C. § 162(q)). Note however that this statute does not include settlements made due to claims of infidelity. This reform avoids the Scylla and Charybdis of enforcing none or all NDAs. Instead, the reform is an intermediate one that literally taxes promises of nondisclosure. And while the tax is nominally borne by the payor (who as we have seen from recent events may be a business or even a perpetrator’s lawyer), the incidence of this tax may ultimately be borne by anyone in privity with the payor—including the survivors themselves (as non-deductibility may lead toward lower payouts). But the tax reform does nothing to distinguish and target serial NDA users for a heightened possibility of investigation. Fox News might still facilitate Bill O’Reilly’s repeat offending by paying hush money to one survivor after another—albeit with a higher price tag.

This Essay instead proposes conditioning the enforcement of NDA provisions on compliance with three requirements that are aimed at exposing repeat offenders to the prospect of heightened scrutiny and potential investigation. The motivation is that these provisions would deter or incapacitate the worst types of repeat offending that, as we have seen, are all too prevalent. But before describing the proposals, let me emphasize that these reform proposals would do almost nothing to deter offenders from committing their first offense. The contestable choice here is to suggest that the legitimate privacy interests predominate with regard to first offenses, but not with regard to offenders who have embarked on a succession of settlements.

NDAs should be enforceable only if they meet particular disclosure, condition, and escrow requirements. Specifically, NDAs should be enforceable only (1) if they explicitly describe the rights which the survivor retains, notwithstanding the NDA, to report the perpetrator’s behavior to the Equal Employment Opportunity Commission (EEOC) and other investigative authorities; (2) if they explicitly make the accuser’s promises to not disclose conditional on the perpetrator not misrepresenting any of the survivor and perpetrator’s past interactions; and (3) if the underlying survivor allegations are deposited in an information escrow that would be released for investigation by the EEOC if another complaint is received against the same perpetrator.

These three proposals are not merely for default NDA provisions that might be waived, but would represent mandatory rules that are easily justified by the negative externality of repeat offending. The survivor and offender in entering into any individual NDA would not be able to contract around these requirements even if they found that doing so would be in their joint, mutual interest, because doing so would place other people at greater risk of being victims of the offender’s future misconduct.

The three parts of this Essay will discuss each of these proposals in more detail and weigh their benefits and potential drawbacks. While the three requirements are proposed as a single reform and might have some beneficial synergies, they might be implemented individually. Since the different interventions likely require, as we will see, different processes to become law, piecemeal enactment might be a natural path forward.

I. The Express Carve-Out Requirement

One way to reduce the hindering effect of NDAs on the ability of agencies to investigate and act against hostile work environments is to require that NDAs explicitly disclose the legal rights that the settling employees retain to contact and cooperate with agency investigations of misconduct. While individual employees can of course settle their claims, it is against public policy to enforce a provision whereby employees trade away the public’s interest in being informed about violations of Title VII. NDAs should be enforceable only if they explicitly carve out from any nondisclosure, non-disparagement, or confidentiality provision that the settling employees can still volunteer potential violations of Title VII to the EEOC and other authorities. The purpose of the carve-out requirement is to explicitly disclose to settling employees that they still retain rights to cooperate with and initiate informing government officials of potential violations. 7 NDAs already “frequently contain an exception for disclosures required by subpoena or court order [but s]ometimes even this is lacking.” Jon Bauer, Buying Witness Silence: Evidence-Suppressing Settlements and Lawyers’ Ethics, 87 Or. L. Rev. 481, 492 n.21 (2008) (citations omitted).

The unenforceability of NDAs in which the settling employee promises not to communicate with the EEOC is already well established. In Equal Employment Opportunity Commission v. Astra USA, Inc., the First Circuit found that “non-assistance covenants which prohibit communication with the EEOC are void as against public policy.” 8 94 F.3d 738, 745 (1st Cir. 1996). The court held not only that settling employees could respond to EEOC subpoenas and requests for information, but that settling employees were free notwithstanding the agreement to contact the EEOC and volunteer information about Title VII violations—rejecting as repressive “a contention that employees who have signed settlement agreements should speak only when spoken to.” 9 Id. The Securities and Exchange Commission has analogously announced that settlement agreements that impede a settlor from communicating with the Commission about possible securities law violations could violate Rule 21F-17 and even lead to the drafting attorney being barred from appearing before the Commission. 10 17 C.F.R. § 240.21F-17(a) (2017). For more discussion of this rule, see Brian Mahoney, SEC Warns In-House Attys Against Whistleblower Contracts, Law360 (Mar. 14, 2014, 5:16 PM EDT),; and Benjamin David Williams, Separation Anxiety: Best Practices for Employee Severance Agreements, 28 Emp. L. Comment., Feb. 2016, at 1, 4.

The EEOC in recent years has aggressively moved to enjoin the enforcement of overbroad NDAs—even some that contain carve-out provisions that allow settling employees to cooperate in agency investigations. 11 See, e.g., Complaint at 2-5, Equal Emp’t Opportunity Comm’n v. CVS Pharmacy, Inc., 70 F. Supp. 3d 937 (N.D. Ill. 2014) (No. 1:14-cv-863), 2014 WL 540344; see also Williams, supra note 10, at 2-4. In an action against Baker & Taylor, the EEOC entered into a consent decree which required the employer to include a more explicit carve-out about employees’ rights to participate in EEOC actions. 12 Consent Decree at 3, Equal Emp’t Opportunity Comm’n v. Baker & Taylor, Inc., No. 1:13-cv-03729 (N.D. Ill. July 10, 2013). NDAs routinely require settling employees to promptly notify the employer if the employee communicates with a public authority (for example, if the employee receives “a subpoena, deposition notice, interview request, or other inquiry related to a civil, criminal, or administrative suit or investigation”). 13 Williams, supra note 10, at 3. These prompt notification requirements are very likely to chill the willingness of settling employees to volunteer information for fear of whistleblower retaliation. It should be against public policy to enforce these provisions and their inclusion should render an NDA unenforceable.

Rendering NDAs without these explicit carve-outs unenforceable may very well reduce the payouts that settling employees receive. But while settling employees are free to trade via settlement their right to be compensated for a claim, these employees cannot trade the public’s interest in having a right to hear from any employee or ex-employee information about civil rights violations. 14 Cf. James Lindgren, Unraveling the Paradox of Blackmail, 84 Colum. L. Rev. 670, 672 (1984) (arguing that blackmail is wrongful for “suppressing the actual or potential interests” of third parties including the public). But see Jennifer Gerarda Brown, Blackmail as Private Justice, 141 U. Pa. L. Rev. 1935, 1936-37 (1993) (arguing that legalizing blackmail based on incriminating information might deter crime and that the public interest in knowing the information should be weighed against such deterrence benefits). Making clear that settling employees have the right to volunteer information and assist in agency investigations is likely to be particularly useful with regard to repeat offenders—because as the number of violations increase, the chance that one survivor will learn about the existence of other claims may induce settling victims who previously opted not to volunteer information to come forward with evidence of employer wrongdoing. 15 NDAs sometimes require settling employees to inform their employer of any related interactions they have with government officials. Courts might deem such disclosures against public policy as well. Carve-outs may even spur some additional disclosure with regard to one-off violations as settling employees who may later feel less exposed to retaliation can see in clear language that they have the right to volunteer the information. 16 Implicitly the carve-out proposal concerns a mandatory “altering rule” governing the necessary method of contracting around the default ability of settling employees to exercise their free speech rights. See Ian Ayres, Regulating Opt-Out: An Economic Theory of Altering Rules, 121 Yale. L.J. 2032, 2036 (2012). To be effective, the altering rule for this carve-out should require that it be written in easily understood language so as to adequately inform settling employees of their continuing reporting right. One might in addition require that the employee separately initial the carve-out provision to more directly call it to the employee’s attention, and it might even be advisable for the provision to reveal the statute of limitations that the agency in question has for bringing the types of actions raised by the underlying allegations.

II. The Condition Requirement

NDA enforceability should also turn on a mandatory condition of exchange. To wit, a settling employee’s nondisclosure promises should be explicitly conditioned on the perpetrator not materially misrepresenting any of the survivor and perpetrator’s past interactions. 17 Under this condition requirement, the parties would not need to stipulate in advance what misrepresentations would qualify as material. Misrepresentations that denied illegal conduct would be material as a matter of law. An accuser later wanting to cancel the contract would be able to seek a court declaration that the accused made a material misrepresentation. Under such a condition of exchange, the settling employee’s nondisclosure promise would remain binding so long as the accused remained silent about their past interactions or spoke truthfully about them. But the settling employee would not be required to stand mute and leave unrebutted misrepresentations by the accused or related parties such as the accused person’s employer or attorney. It should be against public policy to enforce an unconditional nondisclosure promise that is used to let intentional misrepresentations concerning sexual misconduct go unrebutted. 18 The ability to defensively use what would otherwise be confidential information to correct misrepresentations of fact has implicitly been supported in other circumstances. For example, instead of prohibiting recordings of telephone conversations, it might be wiser to prohibit only improper uses of recordings, so as to allow the use of recordings to rebut misrepresentations of what was said in a conversation. See Ian Ayres, Opinion, Why Prosecute Linda Tripp?, N.Y. Times, Aug. 10, 1999, at A17 (“Mrs. Tripp, in her own defense, should have been allowed to use the tapes she made of her conversations with Ms. Lewinsky to prove what was actually said.”). More recently, both Donald Trump and James Comey have seemed to welcome the use of what might otherwise be deemed a confidential recording to rebut a supposed misrepresentation by the other. See Open Hearing with Former FBI Director James Comey: Hearing Before the S. Select Comm. on Intelligence, 115th Cong. 18 (2017) (statement of James Comey, Former Director, Federal Bureau of Investigation) (“Lordy, I hope there are tapes.”); Donald J. Trump (@realDonaldTrump), Twitter (May 12, 2017, 5:26 AM), (“James Comey better hope that there are no ‘tapes’ of our conversations before he starts leaking to the press!”).

This condition proposal is related to an even more basic idea that courts should not enforce promises to lie. A version of this issue has been raised in a recent dispute growing out of an agreement settling a sexual harassment suit that Andrea Mackris, a former Fox News producer, had filed against Bill O’Reilly. The settlement required all parties in the case to disclaim certain evidence “as counterfeit or forgeries” should it be made public. 19 Tom Kludt, Terms of Bill O’Reilly Settlements Revealed for First Time, CNN Money (Apr. 4, 2018, 7:55 PM ET), This provision goes beyond a promise of nondisclosure, and seemed to require Mackris to “lie—even in legal proceedings or under oath—if any evidence becomes public, by calling evidence ‘counterfeit’ or ‘forgeries.'” 20 Id. Of course, promises to lie about sexual misconduct should be unenforceable as against public policy. 21 Ambra Battilana Gutierrez’s confidentiality agreement with Harvey Weinstein provides a particularly egregious example of such a promise. Gutierrez, a victim of his sexual harassment, agreed to a sworn statement “stat[ing] that the behavior Weinstein admits to in the audio tape never happened.” Farrow, supra note 2 (emphasis added). And it is a short step to mandatorily imply a promise that one side will not lie about the sexual misconduct about which the other side has promised to be silent. The condition of exchange doesn’t guarantee that the accused will not falsely deny the sexual misconduct, but it at least gives the accuser a right to rebut the untruth.

The proposal that the “misrepresentation” condition be an explicit part of NDAs only applies to prospective NDAs. But with regard to existing NDAs, courts should find this condition implied as part of the general requirement of good faith. 22 Good faith is often defined to require “honesty in fact”, U.C.C. § 1-201 (Am. Law Inst. & Unif. Law Comm’n 2017), and excludes conduct that “violate[s] community standards of decency, fairness or reasonableness.” 2 Restatement (Second) of Contracts § 205 cmt. a (Am. Law Inst. 1981). But see Ariel Porat & Omri Yadlin, A Welfarist Perspective on Lies, 91 Ind. L.J. 617, 618, 628 (2016) (arguing that if people can be silent about something, they should have the option to lie too, since silence will be interpreted as an equally incriminating reply, but that such unraveling need not occur if particular speakers establish policies not to respond to certain kinds of questions). It is bad faith to bargain for someone else’s silence about certain facts and then to lie about what really happened. For example, falsely denying harassment or adultery should be deemed a material breach of an implicit promise not to lie which gives the settling employee the option to cancel the contract. 23 While my proposal is crafted as a defensive enforcement condition, which conditions one side’s duties to perform on compliance by the other side and therefore “enhance[s] the security of one of the parties by providing incentives for the proper completion of performance or alternatives in its absence,” Steven J. Burton & Eric G. Andersen, The World of a Contract, 75 Iowa L. Rev. 861, 873 (1990), one might imagine that a misrepresentation would also give the settling employee the option of suing for damages.

A form of this argument has been raised in the current NDA dispute between Michael Cohen, President Trump’s longtime personal lawyer, and Stormy Daniels, an adult-film star whose legal name is Stephanie Clifford. Daniels is seeking to void a Cohen-brokered NDA that bars her from speaking publicly about an affair she claims to have had with Trump more than a decade ago. The agreement between Daniels and Cohen called for each party to keep quiet about the incident and the existence of an agreement. Daniels argues that by speaking about the matter in the press, Cohen has breached their contract, and thus she is now excused from performance of the NDA. 24 Dylan Matthews, Stormy Daniels’s Legal Battle Against Trump, Explained, Vox (Mar. 13, 2018, 9:10 AM EDT),

The proposed condition of exchange does not impose reciprocal nondisclosure duties on the settlers. The accused paying hush money might retain the right to comment truthfully on the facts covered by the NDA. But the condition of exchange moves NDAs toward more symmetric duties—as neither side can misrepresent the truth without consequence. The condition would free the settling employee to speak when disclosure is most needed. There is an unfortunate power to unrebutted false denials of sexual wrongdoing. The condition allows the employee to speak truth to power. The exchange is also likely to be especially helpful with repeat offending—because the repeat offender who issues blanket denials concerning sexual misconduct would thereby release all previous NDA survivors from their nondisclosure covenants. 25 After being caught bragging in vulgar terms about kissing, groping, and trying to have sex with women during a recorded 2005 conversation, Trump characterized those comments as “locker room talk” and denied actually behaving that way toward women. Megan Twohey & Michael Barbaro, Two Women Say Donald Trump Touched Them Inappropriately, N.Y. Times (Oct. 12, 2016), Many of his accusers stated that Trump’s denials provoked them into going public with their allegations. See, e.g., Joe Capozzi, Local Woman Says Trump Groped Her, Palm Beach Post (Oct. 12, 2016 7:16 PM),; Natasha Stoynoff, Physically Attacked by Donald Trump—A PEOPLE Writer’s Own Harrowing Story, People (Oct. 12, 2016 10:31 PM),; Twohey & Barbaro, supra.

III. The Escrow Requirement

The final “escrow” proposal is particularly targeted to respond to the problem of “superstar” harassers who engage in repeated sexual misconduct. Just as financial escrows allow parties to deposit money with an escrow agent to be released to pre-specified people upon the occurrence of pre-specified conditions, information escrows are mechanisms that allow deposits of information with an escrow agent to be released to pre-specified people on the occurrence of pre-specified conditions. 26 See Ian Ayres & Cait Unkovic, Information Escrows, 111 Mich. L. Rev. 145, 147 (2012). Information escrows have been used in a wide variety of circumstances—including everything from lawsuit settlement offer escrows, to criminal expungement escrows, to escrow mechanisms in forty-one states that allow adoptees to potentially connect with their biological parents. 27 Id. at 154 (settlement escrows); id. at 152 (expungement escrows); id. at 158 (adoption shared-interest escrows). Indeed, the central feature of the Tinder dating app is an information escrow. When a user swipes right, that indication of interest is not immediately passed on; it is held in escrow by the app and only released if it is matched by reciprocated interest.

The escrow idea has also been deployed by a sexual assault report platform, named Callisto, which is now in use by more than a dozen universities (including Stanford). 28 Callisto, Year 2: Of Combating Sexual Assault, Empowering Survivors, and Advancing Justice; 2016-2017 School Year Report 3, (archived May 13, 2018). I have collaborated with Callisto and its leadership in various capacities for much of its existence. See generally, e.g., Ian Ayres et al., Act-Sampling Bias and the Shrouding of Repeat Offending, 103 Va. L. Rev. Online 94 (2017). In addition to traditional reporting options, the Callisto platform gives survivors a “matching option” of depositing an encrypted, time-stamped complaint into escrow that will be released to the school’s Title IX coordinator for investigation only if another complaint is received accusing the same person. 29 See generally Callisto, supra note 28, at 3 (describing how the school will only be notified if another student reports the same perpetrator). Callisto uses simple game theory to respond to the “first-mover disadvantage” experienced by some survivors. 30 See Ayres & Unkovic, supra note 26, at 147 (describing potential reasons for such a first-mover disadvantage). Survivors who use the matching option don’t have to worry about being the only one to accuse a particular offender, because their accusation will only be released for investigation if there is a second allegation. 31 The Callisto matching option does not need to decrease the number of formal complaints filed. See Ian Ayres, Voluntary Taxation and Beyond: The Promise of Social-Contracting Voting Mechanisms, 19 Am. L. & Econ. Rev. 1, 38-39, 39 fig.3 (2017) (describing a randomized between subject vignette survey showing no reduction in formal reporting caused by offering the Callisto-like matching option). Callisto has now embarked on expanding its offering to use information-escrow matching to aid workers in particular industries who have experienced various forms of sexual misconduct. 32 See What We Do, Callisto, (archived May 13, 2018).

The escrow idea could also help respond to the problem of repeat offenders using a series of NDAs to shield themselves from public inquiry. Specifically, the EEOC might require that whenever an NDA concerning sexual misconduct is agreed to, each party independently deposit their beliefs about the alleged events and copies of relevant evidence into an information escrow mechanism that would be operated by the EEOC or its designate. 33 A version of this escrow requirement might merely acknowledge in the carve-out disclosure that each side has the option of making such an EEOC escrow deposit. Such an optional implementation would more closely parallel the Callisto reporting platform in which the matching escrow is just one of several reporting options. But in circumstances where a settling employee is receiving consideration in exchange for a nondisclosure promise, it may be reasonable as an intermediate intervention targeting repeat offending to go further and require escrow deposits. Crucially, these party submissions would not be vetted by the other side and could not be subject to any agreement as to what allegations would be disclosed—rather, the individual parties at the time of signing the NDA would independently deposit their side’s unfiltered view of the events in a kind of “allegation booth.” 34 Cf. Ian Ayres & Jeremy Bulow, The Donation Booth: Mandating Donor Anonymity to Disrupt the Market for Political Influence, 50 Stan. L. Rev. 837, 838 (1998) (describing the “donation booth” as a variant on secret voting booths in which political donations are funneled through a blind trust, thus keeping candidates from learning donors’ identities). These escrow allegations would be released for investigation by the EEOC (or the Department of Education with regard to Title IX violations) if another complaint was later received naming the same perpetrator. 35 The second, trigger complaint might be also deposited in the NDA escrow mechanism or it might come from a direct complaint to the EEOC.

This escrow proposal would not be a complete solution. It would not deter initial offenses (nor the effective shielding of that offense with NDA hush money). The escrow’s release of information would not be triggered unless and until there was a second report. But the perpetrator’s knowledge that a report of subsequent offending would trigger release of the deposited allegations might deter repeat offending. The escrow NDA device thus represents an intermediate solution between all-or-nothing NDA enforcement choices that targets repeat offending by the superstar offender who can afford a series of NDA payments. It provides some of the privacy benefits for initial survivors but releases the information when the societal interest in deterring repeat offending is greater. The NDA escrow also allows the victim who is uncertain about whether the accused’s wrongful action is characterological of bad behavior to settle the instant complaint while holding the accused potentially accountable if he or she offends again.

The appetite for escrow-like contingent disclosure to spur investigation and potential punishment if the perpetrator continues to offend has been shown in the Weinstein affair itself. It has been reported that Zelda Perkins successfully bargained for provisions in her settlement with Miramax regarding Weinstein’s alleged misconduct that would trigger disclosure to Disney and the firing of Weinstein “if a subsequent sexual-harassment settlement [concerning him] was reached in the following two years.” 36 Farrow, supra note 2.
This escrow proposal is closely related to a previous idea of Samuel Estreicher. 37 See Samuel Estreicher, How To Stop the Next Harvey Weinstein, Bloomberg (Nov. 12, 2017, 7:00 AM PST),; see also N.Y. Univ. Sch. of Law, Avoiding the Next Harvey Weinstein: Sexual Harassment & Non-Disclosure Agreements at 1:00:00-1:02:15, YouTube (Feb. 8, 2018), (comments of Samuel Estreicher). Estreicher has argued that the EEOC should:

require employers to provide data on the number of settlement agreements they have entered into involving allegations against particular employees. In the first instance, identities would not need to be revealed. When a pattern of repeated settlements emerges, the agency could intervene, by opening an investigation or filing a commissioner’s charge (which does not require the accusers themselves to file formal charges). 38 Estreicher, supra note 37.

Estreicher’s proposal is escrow-like, because if the EEOC learns that an employer has entered into multiple settlements with regard to some, as yet, unspecified employee, the EEOC can trigger release of the accused employee’s identity (as well as the details of the multiple accusations) by “opening an investigation or filing a commissioner’s charge.” Because these EEOC actions would not require the accusers to file charges, they would not force an accuser to violate any settlement provision under which the accuser promised not to initiate a proceeding. And as we’ve seen, survivors, notwithstanding any settlement or NDA, always remain free to cooperate with an EEOC investigation.

An important advantage of Estreicher’s proposal over mine is that it might be accomplished without congressional action (although it might require notice-and-comment rulemaking). The existing statutory authorization under Title VII that allows the EEOC to require firms with more than 100 employees to annually disclose the gender and racial composition of their workforce on the EEO-1 form, Estreicher suggests, could authorize the EEOC to collect the NDA information of his proposal. 39 See N.Y. Univ. Sch. of Law, supra note 37, at 1:01:51-1:02:10.

But the Estreicher proposal seems to not require firms to initially disclose the names of the accused employees—rather the employer would merely disclose that, say, five NDAs had been entered into with regard to employee A and three with regard to employee B. The EEOC would not learn that employee A was Bill O’Reilly or Harvey Weinstein. To be effective, Estreicher has suggested that his proposal would also require the EEOC to respond to a disclosure of repeat NDAs by routinely opening investigations seeking additional information about the identity of the employee accused of repeat offending and the nature of the offenses. In contrast, the escrow proposal would automatically release not just the identity of the repeat offender but the nature of the allegations related to each NDA to the EEOC. The escrow proposal would thus have the advantage of proactively giving the EEOC a much richer set of information related to the allegations of repeat offending before the EEOC takes action.


Even when the problem is important and the wrong obvious, it is often not clear which remedial steps are worth the effort or which initiatives should claim our priority. Organizations have limited bandwidth to implement any individual change. Repeat offending by superstar rainmakers that is facilitated by the payment of serial NDA hush money has rightfully claimed our attention, but it is less clear how best to respond. In this Essay, I’ve proposed three potential intermediate legal reforms to “mend, not end” the enforceability of NDAs. Any of these proposals might be implemented by state or federal statute. But enacting statutes, especially in this age of political polarization, is a heavy lift. Thus, it is especially appropriate to consider alternative implementations. The condition proposal might be implemented by judges exercising their common law power to explicate the meaning of good faith. And either the condition or the disclosure proposal might be captured by EEOC guidance viewing noncomplying NDAs as inimical to a non-hostile work environment. 40 In 2012, the EEOC announced new enforcement priorities pledging to “target policies and practices that discourage or prohibit individuals from exercising their rights under employment discrimination statutes, or which impede the EEOC’s investigative or enforcement efforts.” U.S. Equal Emp’t Opportunity Comm’n, Strategic Enforcement Plan FY 2013-2016, at 10 (2012),; cf. Opening Brief of the Equal Employment Opportunity Commission as Appellant at 51, Equal Emp’t Opportunity Comm’n v. CVS Pharmacy, Inc., 809 F.3d 335 (7th Cir. 2015) (No. 14-3653), 2015 WL 2064199 (“This Section 707 case arose because the EEOC had reasonable cause to believe that CVS used the challenged language in its [separation agreement] intentionally to deter workers from filing charges or otherwise communicating with the EEOC.”). The duty to deposit Callisto-like escrow deposits with every sexual misconduct NDA may have a game-theoretic elegance but it is best accomplished by a separate federal statute that isn’t (and given other legislative priorities probably shouldn’t be) likely to occur anytime soon. The Estreicher alternative requires a more proactive EEOC trigger but is probably the closest we can come to an escrow effect.

* William K. Townsend Professor, Yale Law School. Amen Jalal, Greg Klass, Vinay Nayak, and Tracy Nelson provided excellent comments.