Tech

UK drafts strict rules for AI used in hiring

New legislation aims to prevent algorithmic bias

Von ZenNews Editorial 14.05.2026, 20:47 9 Min. Lesezeit

UK drafts strict rules for AI used in hiring

The UK government has moved to introduce sweeping new regulations governing the use of artificial intelligence in recruitment and employment decisions, with draft legislation proposing mandatory transparency requirements, bias audits, and enforceable accountability standards for companies deploying algorithmic hiring tools. The proposals, which have drawn both praise from civil liberties groups and scrutiny from industry bodies, represent one of the most detailed attempts by any government to regulate AI at the point of hiring — a practice now used by an estimated 55 per cent of large UK employers, according to data from the Chartered Institute of Personnel and Development.

What the Draft Rules Propose

The legislation, which is currently in consultation phase, would require employers and third-party vendors supplying AI-driven hiring tools to disclose when automated systems are being used to screen, rank, or reject job candidates. Under the proposals, candidates would have the right to request a human review of any automated decision affecting their application — a provision modelled in part on protections already embedded in the UK General Data Protection Regulation, but extended significantly in scope and enforceability.

Lesen Sie auch

Officials said the rules would apply to any software using machine learning — a form of AI in which systems learn patterns from historical data rather than following explicitly programmed instructions — to make or influence employment decisions. This includes resume-screening tools, video interview analysis platforms that score candidates on facial expressions or speech patterns, and psychometric testing software that generates hiring recommendations automatically.

The draft also proposes that algorithmic tools used in hiring undergo independent bias audits before deployment and at regular intervals thereafter. An algorithmic bias audit involves a structured examination of whether a system's outputs disadvantage particular groups — for example, by producing lower scores for female applicants, candidates from ethnic minority backgrounds, or those with disabilities — relative to outcomes that would be expected under fair, non-discriminatory conditions.

Transparency and Disclosure Requirements

Among the most significant provisions is a requirement that job postings explicitly state when AI tools will be used in the selection process. Employers would be obliged to publish summary information about the logic behind automated decisions and to retain records of how AI systems informed hiring outcomes for a minimum of three years. The Information Commissioner's Office, the UK's data protection authority, is expected to be designated as the primary enforcement body, officials said.

Penalties for Non-Compliance

The proposed penalty framework would allow fines of up to four per cent of global annual turnover for serious breaches — a threshold mirroring the enforcement architecture of the EU's General Data Protection Regulation and consistent with the trajectory of broader AI safety policy tightening ahead of US legislation. Smaller employers, defined as those with fewer than 50 staff, would receive a transitional period before the rules take full effect, though they would not be exempt from disclosure requirements from day one.

Key Data: According to Gartner, more than 60 per cent of large organisations globally are currently using some form of AI or automation in their talent acquisition processes. Research published by MIT Technology Review found that AI-based resume screening tools trained on historical hiring data have, in several documented cases, systematically deprioritised female applicants for technical roles. The UK labour market currently sees over 1.2 million vacancies advertised per month, a significant proportion of which are processed through automated applicant tracking systems, according to the Office for National Statistics.

The Scale of AI Use in UK Hiring

The legislative push comes against a backdrop of rapid and largely unregulated adoption of AI-driven recruitment technology across the British economy. Over the past several years, the cost and accessibility of machine learning software has dropped substantially, placing tools once reserved for large multinationals within reach of mid-sized businesses and even some smaller firms.

According to IDC, the global market for AI-enabled HR technology — encompassing recruitment, performance management, and workforce planning — is currently valued at several billion dollars and is growing at a double-digit annual rate. In the UK specifically, the recruitment technology sector has expanded rapidly, with platforms offering automated video interviews, AI-generated candidate ranking, and real-time psychometric profiling now used by organisations across finance, retail, healthcare, and the public sector.

How Algorithmic Bias Occurs

Algorithmic bias in hiring typically arises in one of two ways. The first is through biased training data: if a machine learning model is trained on historical hiring decisions made by humans who themselves exhibited discriminatory preferences — consciously or otherwise — the model will learn to replicate those preferences. The second is through proxy discrimination, in which an algorithm uses a variable that appears neutral, such as postcode or university attended, but which correlates strongly with a protected characteristic such as race or socioeconomic background.

Wired has previously reported in detail on multiple cases in which AI hiring tools produced racially and gender-skewed results, including the now widely documented case of a major e-commerce company that quietly abandoned an internally developed AI recruitment tool after discovering it was systematically ranking male candidates higher than female ones for software engineering positions. That case, which became a reference point in academic and policy discussions, underscored the risks of deploying machine learning systems without ongoing independent scrutiny.

Industry Response and Concerns

Reaction from the technology and recruitment industries has been mixed. Several HR technology vendors have publicly welcomed the direction of the proposals, arguing that a clear regulatory framework would actually benefit responsible operators by setting a level playing field and preventing a race to the bottom on standards. Others have raised concerns about the practicality of the bias audit requirement, particularly for smaller vendors that may lack the resources to commission independent technical assessments on a recurring basis.

The Recruitment and Employment Confederation, which represents staffing agencies and in-house recruitment teams, said in a statement that while it supported the goal of preventing discriminatory outcomes, it would be seeking clarity from ministers on which specific tools would trigger the audit requirement and whether off-the-shelf software purchased from a vendor would shift compliance responsibility to the vendor, the employer, or both, officials said.

Vendor Liability Questions

The question of where liability sits in a supply chain — particularly when an employer purchases a commercial AI tool and configures it for their own hiring process — is among the most technically and legally complex issues raised by the draft rules. Legal experts quoted in commentary for multiple UK outlets have noted that current employment and data protection law does not cleanly resolve this question, and that the new legislation will need to address it explicitly if it is to be enforceable in practice.

AI Hiring Tool Type	Common Use Case	Known Bias Risk	Covered by Draft Rules
Resume / CV Screening Software	Filtering applications by keyword, experience, or ranking score	Replicates historical hiring bias from training data	Yes — mandatory disclosure and audit
Automated Video Interview Analysis	Scoring candidates on speech, tone, facial expression	Facial recognition accuracy disparities across ethnicities; accent discrimination	Yes — human review right applies
Psychometric / Personality Testing Platforms	Generating personality profiles and suitability scores	Neurodivergent and disability-related outcome disparities	Yes — audit and transparency obligations
Salary and Offer Recommendation Engines	Suggesting compensation bands for candidates	Gender and ethnicity pay gap perpetuation	Partially — under review in consultation
Applicant Tracking Systems (Basic ATS)	Organising and storing applications without automated scoring	Lower risk when no automated scoring is applied	No — excluded if no automated ranking function

The Broader UK AI Regulatory Context

The hiring AI rules do not exist in isolation. They are part of a wider and accelerating shift in UK digital policy toward sector-specific AI governance, following years in which successive governments favoured a lighter-touch, principles-based approach that critics argued left workers and consumers inadequately protected. As reported previously on ZenNewsUK, the government has also moved to impose strict AI safety rules on major technology platforms, and has introduced AI safety provisions under the Digital Bill that extend obligations to a range of high-risk automated systems.

The UK's trajectory is also being shaped by developments on the continent. The EU AI Act, which is now in the process of phased implementation, classifies AI used in employment as high-risk by definition, imposing some of the strictest obligations in the entire regulatory framework. Analysts and policymakers in Westminster have been closely watching how the EU's approach plays out, with some officials indicating privately that the UK will look to maintain a degree of alignment to avoid creating barriers for technology vendors operating across both markets. Coverage of those developments is available in ZenNewsUK's reporting on how the EU has tightened AI regulation with landmark compliance rules.

Divergence From US Approach

While the UK and EU have both moved toward binding regulatory frameworks, the United States has so far relied primarily on a patchwork of state-level legislation and agency guidance rather than federal law. New York City's Local Law 144, which requires bias audits for AI hiring tools used within city limits, is among the most cited examples of targeted municipal regulation, but there is no equivalent federal statute. The contrast with the UK's proposed approach is significant and reflects differing philosophies about the appropriate role of government in technology governance, according to analysis published by MIT Technology Review.

What Comes Next

The consultation period on the draft rules is expected to run for several weeks, after which ministers will review submissions from employers, technology companies, trade unions, civil society organisations, and legal experts before introducing a revised version of the legislation to Parliament. Equality and digital rights advocates have already submitted detailed responses urging the government to strengthen the audit requirements and to ensure that enforcement is adequately resourced — a concern given that the ICO has historically faced criticism for under-enforcement of existing data protection law in employment contexts.

The Trades Union Congress, which represents millions of workers across the UK, has called for unions to be given a formal role in the auditing process, arguing that worker representation in the oversight of hiring algorithms is essential to ensure that compliance assessments reflect the experience of those most affected. Officials have said the government is considering how to incorporate stakeholder input into the audit framework without creating processes so burdensome that they deter adoption of AI tools altogether — a balance that has proven elusive for regulators in other jurisdictions.

For employers currently using AI in recruitment, legal and HR advisers are already recommending that organisations begin mapping which tools they use, how automated decisions are generated, and what documentation currently exists around those processes — steps that will likely be required regardless of the final form the legislation takes. As the UK's regulatory environment for artificial intelligence continues to harden across multiple domains, the hiring rules represent a concrete test of whether government can translate high-level AI ethics principles into enforceable workplace protections. That test is now formally underway.