When staff or agencies use AI tools: employer liability and the reasonable steps defence in Great Britain

In most workplaces the people who actually run AI hiring and management tools are recruiters, HR staff, line managers and outside agencies. When one of them sets a discriminatory filter, writes a biased prompt or acts on a score in a discriminatory way, the question is who answers for it. Section 109 of the Equality Act 2010 sets out when the employer does, and the one defence it has.

The short answer: anything an employee does in the course of employment is treated as also done by the employer, and anything an agent does with the employer's authority is treated as also done by the employer, whether or not the employer knew or approved (s109(1)-(3)). For acts by employees, the employer has a defence if it shows it took all reasonable steps to prevent them (s109(4)).

What section 109 says

• Employees: anything done by a person in the course of their employment must be treated as also done by the employer (s109(1)).
• Agents: anything done by an agent for a principal, with the principal's authority, must be treated as also done by the principal (s109(2)).
• Knowledge does not matter: it does not matter whether the thing is done with the employer's or principal's knowledge or approval (s109(3)).
• The defence: in proceedings about an employee's act, it is a defence for the employer to show it took all reasonable steps to prevent the employee from doing that thing, or anything of that description (s109(4)). The subsection is framed around employees' acts, not agents'.

Where AI tools come in

• Recruiters and HR staff: a recruiter who adds an age filter, or a manager who rejects candidates flagged by a tool for a discriminatory reason, will usually be acting in the course of employment, in which case the act is treated as the employer's.
• Recruitment agencies and other agents: where an agency screens candidates for an employer with its authority, what it does is treated as also done by the employer. See also recruitment agency rules and employment service-providers.
• Acting without the employer knowing: section 109(3) means an employer cannot rely only on not having known how a tool was set up or used.

What "all reasonable steps" can look like for AI tools

The Act does not list the steps; whether the employer did enough depends on the facts. For AI tools, steps an employer can take and record include:

• a written policy on which tools may be used, for what, and which fields and filters are not allowed;
• controls on who can change filters, prompts and cut-offs, with a log of changes;
• training for recruiters and managers on discrimination and on how the tools work;
• checks of outcomes by group where lawful, and a route for staff and candidates to raise concerns;
• acting on problems once found, and keeping a record of what was done.

Keep evidence that the steps were actually taken, not only written down. Section 109(4) is framed around employees' acts; for agencies, clear written instructions on how tools may be used are a sensible step in their own right.

Liability of staff and agents themselves

Employees and agents can also be personally liable where what they did is treated as the employer's act and amounts to a contravention, even if the employer succeeds with the reasonable steps defence (s110(1)-(2)). They are not liable if they reasonably relied on a statement by the employer that what they were doing was lawful (s110(3)), and knowingly or recklessly making a false or misleading statement of that kind is an offence (s110(4)). For more on this and on instructing or helping discrimination, see instructing and aiding discrimination.

For the Act's other workplace rules, see AI and the Equality Act 2010; for time limits and proof, time limits and the burden of proof.

This page covers Great Britain and is general information on the law as at 2 November 2026, not legal advice.