Federal push against DEI-Linked workplace regulation increases
A legal confrontation unfolding in a Denver federal court is drawing close attention far beyond the borders of Colorado — and well beyond the borders of the United States. At its centre is a deceptively simple question: can a state government require companies to ensure that their artificial intelligence tools treat people fairly in hiring and employment? The Trump administration's answer, delivered emphatically this week, is no.
On 24 April, the US Department of Justice filed papers joining a lawsuit originally brought by Elon Musk's AI company, xAI, seeking to prevent Colorado's Anti-Discrimination in AI Act from taking effect. The law, known as SB 24-205, is scheduled to come into force on 30 June. Its objective — preventing algorithmic bias in hiring, housing, healthcare and other high-stakes decisions — might seem uncontroversial in many parts of the world. In the United States of 2026, it has become a political flashpoint.
Read next: How AI is helping Human Resources get back to being human
For HR leaders based outside the United States, the case offers a revealing window into how the current administration is approaching diversity, equity and inclusion — and what that shift may mean for multinationals operating on American soil.
The law at the centre of the storm
Colorado's AI Act was signed into law in May 2024 and made history as the first comprehensive state statute in the US to regulate algorithmic discrimination. Its architecture will be familiar to HR professionals accustomed to the European Union's AI Act: it identifies "high-risk" AI systems — those that meaningfully influence consequential decisions about individuals — and places compliance obligations on both the developers who build them and the organisations that deploy them.
In the employment context, high-risk AI systems include tools used to screen job applicants, rank candidates, evaluate performance, recommend promotions or determine eligibility for workplace benefits. Employers using such tools are required to conduct written impact assessments, maintain documented AI governance programmes, issue transparency notices to affected individuals and report discriminatory outcomes to the state's Attorney General.
The law defines algorithmic discrimination broadly, covering differential treatment on the basis of age, disability, ethnicity, race, religion, sex, veteran status and other protected characteristics. It has no minimum employee threshold — any organisation deploying a covered AI tool in Colorado, regardless of its size, falls within scope.
Enter the federal government
The legal challenge began on 9 April, when xAI filed a complaint in federal court in Denver arguing that the law was unconstitutional on multiple grounds. The company contended that developing an AI model is a form of protected expression under the First Amendment, and that compelling developers to redesign their systems to avoid disparate outcomes amounts to government-mandated speech. xAI also argued that the law was impermissibly vague and would burden commercial activity across state lines.
Two weeks later, the Justice Department moved to join the case — a step that legal analysts described as the first time the federal government has intervened in a lawsuit challenging a state AI regulation. The DOJ's filing added a constitutional dimension centred on the Fourteenth Amendment's Equal Protection Clause, arguing that the law effectively forces AI companies to make decisions on the basis of race, sex and religion in order to satisfy its requirements.
Read next: University of Alberta drops equity hiring preference in rebranding of inclusion strategy
The framing chosen by DOJ officials was striking in its directness. Harmeet K. Dhillon, the Assistant Attorney General for Civil Rights, stated that laws requiring AI companies to incorporate what she called "woke DEI ideology" into their products are illegal. The filing argued the law "jeopardises the United States' position as the global AI leader."
That language was not incidental. It reflects a carefully constructed ideological position that the administration has been building since taking office: that DEI-linked policies — whether in corporate hiring, regulatory compliance or, now, AI governance — represent a form of unlawful discrimination rather than a remedy for it.
A pattern, not an isolated case
For international HR leaders, the Colorado lawsuit is best understood not as a standalone dispute but as one piece of a larger and accelerating federal effort to dismantle the institutional architecture of DEI in American workplaces.
Since January 2025, the Trump administration has issued executive orders dismantling federal DEI programmes, directing agencies to investigate private-sector diversity initiatives, and threatening the federal contracts of companies that maintain certain equity-focused policies. In December 2025, the President signed an executive order specifically identifying Colorado's AI Act by name as an example of state-level regulation embedding ideological bias into technology — the only state AI law singled out in that order.
Read next: DEI goes dark: Employers retreat from transparency, not inclusion
The Justice Department's intervention in the xAI case is a direct extension of that posture. It signals that the administration is prepared to use the courts — not merely executive directives — to challenge state and private-sector frameworks designed to reduce bias and promote fairness in AI-driven employment decisions.
Colorado's law contains a provision that has become central to the federal government's legal argument: it exempts from liability AI tools designed to advance diversity or redress historical discrimination. The DOJ characterises this carveout as an unconstitutional double standard — permitting discrimination in favour of some groups while penalising it against others. It is an argument that legal scholars will debate vigorously, but its political purpose is clear: to reframe DEI-oriented compliance frameworks as themselves discriminatory.
The view from outside the United States
The contrast with developments elsewhere in the world could scarcely be more pronounced.
The European Union's AI Act, which entered its enforcement phases in 2024 and 2025, imposes risk-based obligations on AI systems used in employment that are broadly analogous to Colorado's approach — and far more extensive in their reach. The EU framework requires conformity assessments, human oversight mechanisms and transparency to affected workers, and applies to any organisation deploying covered AI tools within the EU, regardless of where those tools were developed.
The United Kingdom, while pursuing a lighter-touch regulatory model, has maintained that existing equality law applies to AI-assisted employment decisions and has tasked the Equality and Human Rights Commission with monitoring compliance. Canada, Australia and several Asian jurisdictions are at various stages of developing comparable frameworks.
Read next: Nike investigated for 'discrimination against White people'
For multinationals headquartered outside the US, or for international companies with significant American operations, the divergence is creating genuine complexity. An AI screening tool that must meet algorithmic fairness standards in Frankfurt or London may now face legal exposure for attempting to meet equivalent standards in Denver.
HR leaders at global organisations should be aware that what is being contested in the Colorado case is not merely a local compliance question. It is a dispute about whether the concept of algorithmic fairness — using data and testing to ensure AI tools do not systematically disadvantage protected groups — is legally permissible under the US Constitution. The DOJ is arguing, in effect, that it is not.
Practical implications for international HR professionals
Whether or not the June 30 deadline survives the courts, international HR leaders with exposure to the American market face a set of immediate considerations.
Review your AI vendors' US compliance posture. If your organisation uses AI tools from US-based developers for any employment-related decisions — including applicant tracking, skills assessment, performance management or workforce planning — those vendors are now operating in a deeply uncertain legal environment. Understanding how they are responding to both Colorado's law and the federal challenge is essential.
Audit your global AI governance framework for US exposure. Many multinationals have developed global AI ethics or fairness policies that align with EU or UK regulatory expectations. Those frameworks may include provisions — bias testing, impact assessments, diversity-sensitive design — that are now being characterised by the US federal government as legally suspect when applied in an American context. Legal counsel should review whether those provisions create any risk under the emerging US framework.
Read next: Trump orders agencies to end deals with DEI-engaging contractors
Monitor the broader federal direction. The Colorado lawsuit is unlikely to be the last of its kind. The administration has made clear that it views state-level AI regulation as an obstacle to US technological competitiveness, and the DOJ's willingness to enter litigation suggests a sustained strategy. HR leaders should track not only this case but any further executive or legislative action at the federal level that may preempt state AI laws.
Do not abandon compliance preparation for Colorado. Courts move slowly, and an injunction blocking the law before 30 June is not guaranteed. Organisations with Colorado operations should continue preparing for compliance while monitoring the litigation. The law's transparency and impact assessment requirements represent sound governance practice regardless of their legal status in the US.
Prepare for questions from staff and stakeholders. In many countries, employees and institutional investors expect organisations to demonstrate a proactive commitment to fairness in AI-assisted decision-making. The US legal environment is shifting, but reputational expectations — particularly in Europe — are not. HR leaders may find themselves navigating a gap between what US law increasingly permits and what global stakeholders demand.
What comes next
The case, xAI v. Weiser, is proceeding in the US District Court for the District of Colorado. The law remains scheduled to take effect on 30 June unless a court issues an injunction. The Colorado legislature has until 13 May to pass amendments that could reduce the law's compliance burden — though whether any amendments would satisfy the federal government's constitutional objections is far from clear.
Beyond the immediate litigation, the broader question is whether the United States will develop a coherent national approach to AI in the workplace, or whether the field will remain a patchwork of state laws — some protective, some permissive — contested by a federal administration that has made dismantling DEI-adjacent regulation a central priority.
For international HR professionals accustomed to operating within predictable regulatory frameworks, the uncertainty is itself the message. The United States is engaged in a fundamental argument about what fairness in the workplace means — and who gets to define it.
The case is xAI v. Weiser, 1:26-cv-01515, US District Court, District of Colorado.
This article is provided for informational purposes only and does not constitute legal advice. Organisations should seek qualified counsel in all relevant jurisdictions.
Global AI Hiring Laws: Country Comparison
Status of laws and regulations affecting the use of AI in employment decisions — April 2026
| Jurisdiction | Status | Enforcing body | Key HR obligations |
|---|---|---|---|
| Europe | |||
 European Union |
Binding | National market surveillance authorities | AI Act classifies almost all HR AI as “high-risk.” Requires conformity assessments, risk management documentation, human oversight mechanisms and transparency to workers before deployment. Applies to any organisation deploying covered AI within the EU, regardless of where the tool was built.
High-risk employment provisions phasing in; full enforcement no later than December 2027 |
 United Kingdom |
Binding | Information Commissioner’s Office (ICO) | Data Use and Access Act 2025 (DUAA): impact assessments required for significant automated employment decisions; individuals retain right to human review and override. ICO codes of practice on AI hiring tools updated. Existing equality law already applies to AI-assisted decisions.
Lighter touch than EU AI Act but binding obligations in force |
 Germany & EU states |
Binding | Works councils / labour courts | Works council co-determination rights when introducing AI-based HR tools. Employers must consult and agree with works councils before deploying AI in hiring or performance management. Applies in Germany, France, Spain, Italy, Austria and the Netherlands.
Operates alongside EU AI Act obligations; consent required before deployment |
| North America | |||
 USA — Federal |
Retreating | DOJ / EEOC (limited) | No comprehensive federal AI employment law exists. The Trump administration is actively challenging state laws it regards as DEI-linked regulation. Colorado’s AI Act was the only state law named in a December 2025 executive order. A National AI Legislative Framework was published in March 2026 but is advisory only and imposes no obligations on employers.
Diverging sharply from the direction of most other major economies |
| Colorado (USA) |
In dispute | State Attorney General | SB 24-205 (due 30 June 2026): written impact assessments; AI governance programme; bias monitoring; transparency notices to applicants; reporting of discriminatory outcomes to the Attorney General. Most comprehensive US state AI law to date.
Currently subject to federal lawsuit filed by DOJ and xAI; legal status uncertain |
| Illinois (USA) |
Binding | Illinois Dept of Human Rights | Human Rights Act amendment (in force January 2026): prohibits use of AI that causes discriminatory effects on any protected class in hiring, promotion or termination; advance notice required when AI is used; applies to any employer with one or more employees in Illinois.
In force since 1 January 2026 |
| New York City (USA) |
Binding | NYC Dept of Consumer & Worker Protection | Local Law 144 (in force since July 2023): annual independent bias audit of automated employment decision tools (AEDTs); public summary of audit results required; candidate notice required when an AEDT is used in hiring or promotion.
First US jurisdiction to require bias audits; in force since 2023 |
| California (USA) |
Binding | CA Civil Rights Dept / CPPA | FEHA regulations (in force 2025–2026): AI tools causing discrimination on protected characteristics are unlawful; four-year record retention required. CCPA automated decision-making regulations (January 2026): risk assessments for sensitive data; opt-out rights for significant employment decisions including hiring, firing and performance evaluation.
Multiple overlapping frameworks; among the most detailed in the US |
 Canada |
Partial | Federal / Provincial regulators | Ontario Working for Workers Act (in force January 2026): employers must disclose when AI is used in the hiring process. Federal Artificial Intelligence and Data Act (AIDA) pending: would regulate high-impact AI in employment nationwide. Existing federal privacy law (PIPEDA) applies to AI data processing.
Patchwork of provincial obligations; comprehensive federal law still pending |
| Asia-Pacific | |||
 South Korea |
Binding | National AI Committee | AI Basic Act (in force January 2026): high-impact AI used in employment must allow meaningful human oversight and intervention at all times; transparency and labelling required; risk assessments must be documented; foreign companies must appoint a Korean local representative.
First comprehensive national AI law in Asia-Pacific; closest to EU AI Act in structure |
 China |
Binding | Cyberspace Administration of China (CAC) | Algorithm Recommendation Regulation (in force March 2022): transparency in how algorithms process and recommend decisions. Mandatory labelling rules for AI-generated content. Security assessments required for publicly deployed AI models. Strictest active enforcement regime in APAC.
Multiple binding instruments; enforcement is active and penalties are meaningful |
 Japan |
Promotional | AI Strategic Headquarters | AI Promotion Act (in force June 2025): a promotional framework encouraging responsible AI use; no penalties for non-compliance. Sector-specific guidelines expected throughout 2026. Existing Labour Standards Act and Personal Information Protection Act (APPI) continue to apply to AI-assisted employment decisions.
Innovation-first approach; penalties deliberately excluded from the framework |
 Singapore |
Voluntary + pending | PDPC / Ministry of Manpower | AI Verify toolkit and Model AI Governance Framework: voluntary testing and accountability standards widely adopted. Workplace Fairness Act (passed 2025): Singapore’s first statutory anti-discrimination law; prohibits adverse employment decisions based on protected characteristics; mandatory grievance procedures required.
Workplace Fairness Act expected in force end-2027; AI Verify framework voluntary |
 Australia |
Emerging | OAIC / Fair Work Commission | Mandatory AI guardrails for high-risk uses including employment are proposed but currently recommendatory only. Privacy Act amendment (in force December 2026): individuals gain the right to request information about substantially automated decisions affecting them. Existing anti-discrimination law already applies to AI-assisted hiring decisions.
Binding guardrails under development; watch for December 2026 Privacy Act changes |
| Latin America | |||
 Peru |
Binding | Secretariat for Government & Digital Transformation (SGTD) | Law 31814 and Supreme Decree 115-2025-PCM (in force January 2026): AI used in recruitment, evaluation, hiring and dismissal is classified as high-risk; requires human oversight, algorithmic transparency and documented risk assessments before deployment. Sectoral compliance deadlines phase in from September 2026 (health, education, finance) to September 2029 (other sectors).
First binding AI framework in Latin America; modelled on OECD and EU principles |
 Chile |
Partial / pending | Financial Market Commission / CMF | Personal Data Protection Law No. 21,719 (in force December 2026): individuals have the right to object to decisions based solely on automated processing, including employment decisions — closely aligned with GDPR Article 22. Separate comprehensive AI legislation under discussion.
Data protection rights binding from December 2026; broader AI law pending |
 Brazil /  Colombia /  Mexico |
Draft | TBD | All three countries have AI legislation under active discussion. Brazil’s draft proposes a risk-based framework modelled on the EU AI Act, including the right to an explanation for AI-driven hiring decisions. Colombia and Mexico are at earlier stages of legislative development.
None yet enacted; watch for 2026–2027 developments in Brazil |
