FEATURED TOPIC | MARKET
The Truth About Hate Speech Rules
How broad speech rules can restrict religious expression, political disagreement, and equal access to services.
6-minute read • Market • Updated July 2026
Companies should prohibit threats, stalking, targeted harassment, and incitement to violence. The problem begins when terms of service or acceptable-use policies restrict speech because someone considers it hateful, intolerant, offensive, bigoted, demeaning, or harmful. Those labels often lack objective boundaries and can be enforced differently depending on who speaks, who objects, or which viewpoint is unpopular. |
Why It Matters
These policies can govern banking, payments, cloud services, marketplaces, advertising, hosting, and communications tools. When “offense” becomes the standard, users may lose access for expressing lawful religious, political, or social views rather than for causing a defined harm.
How Hate-Speech Rules Can Restrict Religious Expression
Many faiths make claims about marriage, sexuality, human identity, sin, and religious truth. Those beliefs may conflict with prevailing cultural views. Overbroad policies can easily be interpreted to prohibit:
- Quoting a religious text about sexuality or marriage
- Expressing a traditional understanding of male and female
- Declining to use language that communicates agreement with a disputed belief about gender
- Teaching that certain conduct is sinful or that one faith makes exclusive truth claims
- Criticizing a religion, secular ideology, law, or social movement
Context matters. Repeatedly directing hostile messages at an individual may be harassment. A credible threat is harmful conduct. But expressing a belief on a disputed religious or social question should not become punishable merely because someone objects to it. Policies should distinguish criticism of an idea from harmful conduct directed at a person.
What These Policies Look Like in Practice
The examples below show different drafting problems. Quoting them does not prove that a company enforced the rule against a particular viewpoint. It shows the standards the company publicly reserved the right to apply.
Airbnb: subjective and compelled-language rules |
Airbnb bars language that “demeans, insults, stereotypes, or seeks to convey a person’s inferiority” because of a protected characteristic and specifically identifies deadnaming, misgendering, microaggressions, and “all other forms of hateful speech.” |
Why it matters: Terms such as “demeans” and “microaggressions” can depend heavily on the listener’s interpretation. Deadnaming and misgendering rules can also reach disputed beliefs about sex, gender identity, and compelled language, as well as religious beliefs. |
_Nondiscrimination-Policy-Airbnb-Help-Center_Discriminatory-Language.png){kind=link}
Akamai and Twilio: unequal protection by design |
Akamai’s Linode policy and Twilio’s community code state that moderators prioritize “marginalized people’s safety over privileged people’s comfort” and will not act on certain complaints involving “reverse racism,” “reverse sexism,” or “cisphobia.” |
Why it matters: These policies do more than use vague terms. They announce that some complaints will be treated differently according to group classifications. A neutral rule against harassment should protect everyone from the same harmful behavior. |
Block: a one-way discrimination standard |
Square defines hate and discrimination by reference to protected classes but states that promotion of historically excluded or disadvantaged groups “may not constitute discrimination.” |
Why it matters: Similar conduct may be classified differently depending on which group benefits. A stronger policy would prohibit unfair exclusion, threats, and targeted harassment consistently. |
_Understanding-Squares-Terms-of-Service(etc)_Review-Squares-hate-prohibition-terms.png){kind=link}
Morgan Stanley: financial restrictions tied to offense |
Morgan Stanley’s E*TRADE bill-pay terms prohibit communications and payments involving goods or services that “promote hate,” “racial intolerance,” or language that is “bigoted,” “hateful,” “racially offensive,” “indecent,” or “discourteous.” |
Why it matters: Because this governs a financial service, vague terms can affect lawful books, events, organizations, advocacy, or religious materials. Financial rules should focus on fraud, illegality, sanctions, threats, or other defined risks. |
_Bill-Pay-Terms-and-Conditions_Acceptable-Use.png){kind=link}
Zoom: protection limited to listed groups |
Zoom prohibits threats or harassment based on listed characteristics and reserves the right to add categories. |
Why it matters: Threats and harassment should be prohibited against any person. A list-based rule may leave users without equal protection when the same conduct targets political affiliation, ideology, occupation, or another unlisted characteristic. |
_Acceptable-Use-Guidelines_Hateful-conduct.png){kind=link}
Why Scrutinizing These Terms Is Not Bigotry
Companies often receive lower scores when policies rely on terms such as “racist,” “bigoted,” “hateful,” “intolerant,” “deadnaming,” or “misgendering” without clearly defining the prohibited conduct or applying the standard equally.
That does not mean the Viewpoint Diversity Score endorses racism, hostility, or mistreatment. It means the Index distinguishes between condemning harmful conduct and granting companies broad discretion to punish lawful expression under vague or identity-specific standards.
The analysis asks whether employees or users can understand the rule in advance, whether comparable conduct is treated consistently, and whether the policy targets objective harms such as threats, harassment, or discrimination rather than disagreement with a religious, political, or moral viewpoint.
A lower score, therefore, reflects concerns about clarity, equal application, and viewpoint neutrality. It is not a defense of bigotry or an objection to protecting people from genuine mistreatment.
Harmful Conduct Is Not the Same as Offensive Expression
Companies can address genuine harm through clear, conduct-based rules. Strong policies prohibit:
- Credible threats or incitement to imminent violence
- Targeted and repeated harassment
- Stalking or doxing
- Fraud, impersonation, or unauthorized access
- Unlawful discrimination in a commercial transaction
These rules identify observable conduct and a recognizable harm. By contrast, terms such as “demeaning,” “intolerant,” “discourteous,” and “offensive” often depend on a listener’s reaction. Two people may interpret the same religious or political statement differently.
What Better Policies Look Like
Better approach | How to implement |
Prohibit conduct, not disagreement | Define threats, stalking, repeated targeting, fraud, or another specific action. |
Protect everyone | Apply the same rule regardless of the target’s identity or viewpoint. |
Protect the discussion of ideas | Allow criticism of beliefs, institutions, laws, religions, and social movements. |
Consider context | Account for religious teaching, news, satire, political debate, and academic discussion. |
Limit enforcement discretion | Give users fair notice and explain why enforcement occurred. |
How the Viewpoint Diversity Score Evaluates These Rules
The Harmful Conduct Policies Apply Equally benchmark asks two questions:
- Does the company avoid restricting speech merely because someone may find it offensive?
- Does the company protect everyone from harmful conduct rather than only selected groups?
The Score does not object to rules against threats, bullying, or harassment. It evaluates whether those rules are clear, conduct-based, and equally applied to all.
Frequently Asked Questions
Does the Viewpoint Diversity Score defend hate speech?
It supports lawful expression under the First Amendment, even if some might consider the speech hateful. Speech does not lose protection simply because someone considers it hateful or offensive. The Index allows for clear restrictions on threats, targeted harassment, incitement, and other defined harms while questioning policies that restrict lawful expression based on subjective offense or viewpoint.
Is every harassment rule problematic?
No. A rule can be appropriate when it defines repeated or targeted conduct and clearly explains the criteria that must be met to constitute harassment.
Why examine deadnaming and misgendering rules?
Because they often regulate speech on disputed questions of sex, gender identity, and compelled terminology. Policies should not punish individuals for expressing beliefs about gender identity and biological sex, nor should they compel speech about the same.
Are protected-class policies always unequal?
No. Nondiscrimination rules may be appropriate or legally required in commercial transactions. The concern arises when harmful-conduct protections are available only to listed groups, or when identical conduct receives different treatment based on group identity.
___________________________________
Editorial note: Company examples are drawn from publicly available corporate reports, policies, career pages, regulatory filings, and verified reporting reviewed for the 2026 Viewpoint Diversity Score.
Disclaimer: The information contained in this document is general in nature and is not intended to provide, or be a substitute for, legal analysis, legal advice, or consultation with appropriate legal counsel. You should not act or rely on information contained in this document without seeking appropriate professional advice. By printing and distributing this document, Alliance Defending Freedom is not providing legal advice, and the use of this document is not intended to constitute advertising or solicitation and does not create an attorney-client relationship between you and Alliance Defending Freedom or between you and any Alliance Defending Freedom employee.
Back to Resources