What “Weakened NDA” Actually Means
The subject here is still the response of venture-capital-backed startups to harassment-related NDA reforms, but the focus is narrower: what, exactly, has been weakened? The answer matters because “NDA ban” suggests that all confidentiality disappears, while the law targets a more specific barrier to speaking about sexual assault and sexual harassment. Understanding that boundary lets us judge the startup hiring evidence without confusing legal enforceability with automatic disclosure or assuming that every firm faces the same new risk.
Two different kinds of confidentiality
A nondisclosure agreement is a contract that limits what one party may reveal or discuss. In a technology startup, that can serve an ordinary and legitimate purpose. A new employee may learn a product design, source of revenue, customer list, research result, or other proprietary information. If the firm could not protect such information, it would be harder to share knowledge internally, collaborate with outside partners, or maintain an advantage over competitors. The Speak Out Act explicitly preserves an employer’s ability to protect trade secrets and proprietary information.
The controversy arises when a broad confidentiality promise also reaches workplace misconduct. A clause can be written or interpreted in a way that discourages a worker from discussing harassment, assault, discrimination, or retaliation. The contractual language may not say “protect the perpetrator.” Its practical effect can still be to keep a worker from telling colleagues, a journalist, an investigator, or a future employee what happened. If no one else can see a pattern, the organization can treat each incident as an isolated dispute rather than a signal about its culture or leadership.
The legal and organizational distinction is therefore not between confidentiality and no confidentiality. It is between information that a firm has a legitimate reason to protect and a restriction that prevents people from discussing unlawful conduct. The two can occupy the same employment agreement, which is why a reform aimed at harassment-related silence does not require abandoning trade-secret protection.
Timing creates a second distinction. A predispute NDA is agreed to before a dispute exists, often as a condition of employment. It tells a worker in advance what the firm expects them not to disclose if a problem later occurs. A post-dispute NDA may be part of a settlement reached after an allegation or claim. In that arrangement, confidentiality can be exchanged for compensation or another resolution, and the worker may weigh the immediate benefit of settlement against the future cost of remaining silent.
These arrangements do different institutional work. A predispute clause operates before anyone knows whether an incident will happen. It can shape a worker’s expectations about whether speaking will be possible and can shape a firm’s expectations about how much information will become public. A post-dispute clause operates after the organization has already learned about a conflict. It can limit what other workers learn from the resolution. The state reforms described in the startup study did not all address the same point in this sequence: some restricted settlement confidentiality, while others limited the use of NDAs as a condition of employment.
The statute’s related concept of a nondisparagement clause is also narrower than a general promise to be polite. It concerns a contractual restriction on making a negative statement about another party when the statement relates to the contract, claim, or case. A nondisclosure clause restricts disclosure or discussion of conduct, a settlement involving conduct, or information covered by the agreement. Both can matter when a worker wants to describe harassment, but they are not interchangeable terms. Keeping them separate prevents a reader from treating every employment confidentiality provision as if it had the same legal effect.
A final distinction is between a legal right and a social reality. If a court cannot enforce a clause, that removes one formal obstacle. It does not guarantee that a worker will report, that an employer will listen, or that speaking will be free of retaliation. A person may still fear losing income, being labeled difficult, damaging a career, or being forced to continue working under the same manager. The law changes the available leverage; it does not by itself create trust.
What the Speak Out Act changes—and what it does not
The federal Speak Out Act, enacted on December 7, 2022, provides a clear illustration of this targeted approach. For a dispute involving sexual assault or sexual harassment, a nondisclosure or nondisparagement clause agreed to before the dispute arises is not judicially enforceable when the alleged conduct violates federal, tribal, or state law. The Act applies to a claim filed under one of those bodies of law on or after enactment.
Several boundaries are built into that rule. It concerns covered sexual-assault and sexual-harassment disputes, not every disagreement between an employer and employee. It addresses judicial enforceability, not whether a clause can still influence behavior informally. It preserves state or local rules that are at least as protective of an individual’s ability to speak. It does not displace rules about pseudonyms in filings. And it says that employers and employees may continue to protect trade secrets and proprietary information.
In other words, the federal rule removes a specific legal shelter for predispute silence in a covered dispute. It does not transform every internal document into public information. It does not make an allegation a proven fact. It does not tell a court that an employment relationship contained harassment. It changes what a firm can expect to keep confidential if a legally covered dispute later arises.
That expectation is the link to startup hiring. Imagine two otherwise similar young firms considering how to build a team. Before a reform, both may believe that a future harassment dispute could be constrained by a predispute clause, even if neither firm has experienced such a dispute. After the reform, that assumption becomes less reliable for covered conduct. The first firm may respond by investing in training, an independent reporting route, investigation capacity, and clear consequences. The second may decide that the safest immediate way to reduce exposure is to avoid hiring into relationships it associates with higher risk. This is an illustrative incentive sequence, not a claim that every manager consciously performs that calculation.
The startup study treats a series of state-level laws adopted from 2018 onward as the earlier legal changes and the federal Act as a later extension of the same broad policy direction. The state rules were not identical. Some dealt primarily with settlement clauses; others limited employers’ ability to require harassment-related NDAs as a condition of employment. Their common feature was to weaken the enforceability of agreements that suppress disclosure of harassment-related conduct. Because the details varied, the study’s treatment is best understood as legal exposure to a family of reforms, not as exposure to one uniform contract rule.
That variation is useful for the study but important for interpretation. When states adopt reforms at different times, researchers can compare the change in startup outcomes after a state’s reform with outcomes in places that have not yet adopted one. But legal adoption is not random. Reforming states may also have stronger #MeToo mobilization, more worker-protective policies, and different labor-market conditions. The study therefore needs tests that show whether the hiring change appears at the reform date and in the jurisdiction where the law applies, rather than merely tracking a general rise in public attention.
The federal law also clarifies why the mechanism can be active before an incident appears in a dataset. The relevant economic object is not an observed complaint; it is expected exposure. A firm may revise its estimate of legal, reputational, and investor consequences because the information environment has changed. If a future report can no longer be contained in the same way, the firm may perceive the expected cost of a failure as higher. That perception can affect hiring, internal oversight, or financing even when no researcher can observe the underlying event.
This does not imply that disclosure is the problem. The policy rationale in the statute is that silence can shield perpetrators and allow illegal conduct to continue. If employees can speak, patterns may become visible, decision-makers may face accountability, and future workers may make better-informed choices. Those are the intended benefits. The empirical question is whether organizations respond to the new visibility by improving their capacity or by reducing the number of people whose presence creates perceived exposure.
The distinction is especially consequential in a startup because the firm’s legal and cultural infrastructure is being built at the same time as its workforce. A larger organization may have an established HR function, multiple reporting channels, investigation procedures, and legal counsel. A founder-led startup may rely on a manager who is also a close friend of the person being reported, or on an informal promise that “we will handle it.” The law does not determine which structure exists. It makes the weakness of the structure more expensive if a dispute becomes public or litigated.
The worker-centered evidence in the supplied PLOS One study helps specify what an alternative to silence would need to look like. In interviews with 30 people who had experienced workplace sexual harassment, participants described clear expectations from leaders as a preventive signal. When an incident occurred, they valued an independent or structurally separate person or body to whom they could report, acknowledgment that their account was taken seriously, and follow-up actions that protected them and addressed the conduct. The study is qualitative and concentrated in one region and mostly service-sector work, so it cannot tell us how common these practices are in startups. It does show why “the clause is unenforceable” and “the workplace is safe to speak in” are different propositions.
The same point appears in the abstract of a recent Human Resource Management study, which treats organizational silence as more than an employee’s decision to keep quiet. Its “signals of silence” include being silent, silencing others, and not listening. Ethical supervision is presented as a possible counter-signal. Applied cautiously here, the concept suggests that a formal contract is only one layer of an information system. A startup can remove a restrictive clause yet continue to communicate, through dismissive responses or social retaliation, that disclosure will not lead to protection. Conversely, a credible response system can make a legal right usable rather than merely theoretical.
The sequence has two deliberate limits. First, a worker’s ability to report or disclose is not the same as a decision to do so; fear, retaliation, and distrust can remain. Second, the post-dispute branch is not a claim that the federal Act makes every settlement clause unenforceable. The federal rule described here concerns predispute clauses, while state law and the facts of a dispute can matter for other arrangements. What the law changes most clearly is the reliability of predispute silence as a tool for containing a covered claim.
That is enough to explain how the legal mechanism could reach hiring. It does not yet tell us whether the observed startup result is a true reform-linked response, a shift in public awareness, or a different change happening at the same time. To answer that, we need to examine how the researchers compare treated and untreated startups—and what the data can actually see.