The AI Bill Project

Artificial Intelligence and algorithmic management   

AI is when computers carry out tasks you would usually expect to be carried out by a human. At work, this might include making important decisions about people, like whether they get a job, how they do their work, where they do it, and whether they are rewarded, disciplined or made redundant.   

Algorithmic management (when AI carries out the role of a human manager) has a significant impact on workers, including work intensification, risk to health and safety, discriminatory and unfair outcomes, lack of control over data, loss of privacy, lack of human agency and freedom over work, and the degradation of human judgment and skill.   

TUC AI Project  

The TUC has been conducting a four-year project on the use of AI in the employment relationship, assisted by a trade union working group (The TUC AI Working Group)2 . This project has focused on algorithmic management. We have undertaken research, commissioned a legal report, published a manifesto for change and guidance for trade union reps. All of the reports are here: TUC AI reports. 

There are provisions in existing legislation that provide some helpful tools when dealing with technology at work. These include rights under the Equality Act 2010, the UK General Data Protection Regulation, the Information and Consultation Regulations, health and safety legislation, and the European Convention of Human Rights.  

However, there are significant gaps in the existing legal framework. 

These gaps (identified by Robin Allen KC and Dee Masters in their TUC AI Project legal report AI Managing People: the Legal Implications Legal report) include a lack of transparency and explainability, a lack of protection against discriminatory algorithms, an imbalance of power over data, and lack of worker voice and consultation.   

In response to the urgent and growing impact of algorithmic management, in 2023 the TUC AI Project launched a taskforce to draft an AI Bill, building on the policy positions and values articulated in the TUC’s AI manifesto, Dignity at Work and the AI Revolution AI Manifesto. 3  

The AI Taskforce and Special Advisory Committee  

In September 2023, the TUC set up a taskforce with the AI Law Consultancy at Cloisters Chambers, and the Cambridge University Minderoo Centre for Technology and Democracy, to manage the drafting of the Bill.  

The Bill was drafted by Robin Allen KC and Dee Masters (and colleagues) of the AI Law Consultancy AI Law Consultancy at Cloisters chambers. The TUC was assisted in administration of the project by the Cambridge University Minderoo Centre for Technology and Democracy. 

The Bill benefitted from the input and expertise of a Special Advisory Committee, which met three times across 2023 and 2024. The Committee was chaired by Kate Bell, assistant general secretary of the TUC, and Gina Neff, executive director of the Cambridge University Minderoo Centre for Technology and Democracy.4  

Members of the Committee included representatives from a diverse range of organisations and political parties, including the Ada Lovelace Institute, the Alan Turing Institute, Connected By Data, TechUK, UKBlackTech, the Institute for the Future of Work, the British Computer Society, CIPD, the RAI UK, Cambridge University, Oxford University, Prospect, Community, CWU/UTAW, USDAW, GMB and cross-party MPs.  

The policy expressed in the Bill is that of the TUC and should not be taken to express the policy of these organisations unless explicitly stated. 

All trade unions in the TUC AI Working Group were also given an opportunity to contribute to the Bill, being sent all paperwork during the cycle of input described below, and given the chance to contribute comments in writing.  

The drafting process 

This project was based on the importance of collaborative and multi-stakeholder work in the field of AI policy development.  

The taskforce and SAC met three times at Congress House, Great Russell Street, London: 21st September 2023, 23rd November 2023 and 6th February 2024.  

Before each meeting, the SAC was sent the most recent version of the draft Bill. The SAC was invited to submit written comments to the taskforce before each meeting. Oral comments from the meetings were also taken into account by the taskforce. 

The Bill was amended by the AI Law Consultancy after each meeting, and then recirculated for comment. This cycle continued until the final draft of the Bill was produced on 20th February 2024. 

This has been a genuinely collaborative endeavour, during which the taskforce, in discussion with the SAC, sought to draft pragmatic Bill provisions. 

On 5th March 2024, the Executive Committee of the TUC approved the Bill for publishing. 

On the 18th April 2024 the TUC published the Bill on its website and launched the Bill with an evening event at Congress House, Russell Street, London. 

The AI Bill: key provisions 

AI decision making  

The Bill deals with decision-making at work that is based on artificial intelligence systems. It provides protections and rights for workers, employees, jobseekers and trade unions, as well as obligations for employers and prospective employers.  

AI operates as a system. This diagram, taken from the TUC publication People Powered Technology 5 , sets out the key components of an algorithmic management system. Technology driven surveillance amasses data which is then used to inform fully automated decisions about workers, or other outcomes such as ratings and behaviour nudges. The Bill provides for the fair and safe operation of AI systems.  

• 2 The TUC AI Working Group has representatives from 21 affiliate TUC unions, led by TUC policy officer Mary Towers.
• 3 The project was also inspired by a Private Member’s Bill on AI and Employment sponsored by Mick Whitley MP and based on the TUC’s AI manifesto https://bills.parliament.uk/bills/3464
• 4 Tom Lacey, Cambridge University Minderoo Centre for Technology and Democracy, Secretariat, Mary Towers TUC Deputy Chair and Secretary.
• 5 People Powered Technology: https://www.tuc.org.uk/sites/default/files/2022-08/People-Powered_Techn…

The Bill applies to “high-risk” decision making. This risk-based approach is similar to the one taken by the European Union in their AI Act, and is also based on the TUC’s AI Manifesto.   

The Bill defines high risk being where there are “legal effects or other similarly significant effects”. The Bill leaves scope for more guidance on this, in consultation unions and others.  

Transparency, observability and explainability  

These provisions are at the heart of the Bill, intended to ensure only safe AI systems make it into the workplace, and that workers and unions are fully consulted, involved and informed. As far as possible, the rights are based on existing legal frameworks, but some of this is also brand new and groundbreaking.  

Workplace AI Risk Assessments (“WAIRA”)  

The employer is obliged to carry out detailed AI risk assessments of AI decision making, pre and post deployment, with rolling 12 monthly reviews.  

Consultation  

The WAIRA framework also establishes a consultation process with workers and employees.  

In addition, trade unions have a statutory right to consultation before the deployment of high-risk AI decision making systems in relation to employees, as well as on a rolling 12-month basis.  This statutory right mirrors the existing consultation obligations for employees and their unions in relation to collective consultation and redundancy (s 188 TULRCA 1992).   

Algorithmic transparency  

The WAIRA framework also provides for significant access to information (for workers, employees and unions) about how the AI system is operating – breaking open the “black box” of these types of systems so that unions and workers are in a better position to understand how they operate.  

Register  

There is an obligation for employers to publish a register of the AI decision- making systems in operation.  

Personalised statement  

Employees, workers and job seekers have a right to a personalised statement explaining how AI was making high-risk decisions about them.  

Right to human review  

Employees, workers and job seekers would also have a right to human review of AI decision making.  

Redressing imbalance of power over data  

Redressing the imbalance of power over data at work is one of the key measures needed to ensure that workers have influence over AI at work.  

Union right to data  

There is a right for unions to be given data about union members that is being used in relation to all AI decision making in the workplace. This would go a long way to redressing the current imbalance of power over data at work.  

In addition, ACAS would be obliged to prepare guidance on AI and data at work, addressing a current information and education gap on data at work.  

Red line over emotion recognition tech  

There is a ban on emotion recognition technology which is used to the detriment of workers, employees, and jobseekers.     

Protections against discriminatory algorithms  

The Bill takes steps to guard against the potential for AI to entrench inequality and carry out discriminatory decision making at work.  

Burden of proof reversed 

The Bill reverses the burden of proof, making it easier to prove AI based discrimination has taken place under the Equality Act 2010.  

Auditing and defence  

There is also a new defence to a discrimination claim where employers can show they have properly audited an AI system.   

The Bill provides for auditing AI systems for discrimination. The new defence will act as a strong incentive for employers to carry out the audits.  

Guidance  

In addition, the EHRC is required to publish guidance for employers on how to avoid AI driven discrimination at work –  in consultation with unions and others, and with regular review to ensure it stays up to date with new technologies.  

There is provision in the Bill for further development of general concepts and ideas through guidance from organisations such as ACAS and the ICO.  

A right to disconnect  

To guard against AI driven work intensification, there is a right to disconnect for employees. This right has been drafted based on precedents from Europe and Australia.  

There is a statutory right for employees to disconnect outside of agreed working hours.  

However, this can be varied by collective or workforce agreements.   

ACAS will be required to give guidance on the right to disconnect, in consultation with others including unions.   

And the employer is obliged to provide a statement of the right to disconnect to employees, consulting with unions and taking into account ACAS guidance.   

Right not to be unfairly dismissed by an AI system    

We propose to extend unfair dismissal law to give employees protection against unfair dismissal by AI.  

Regulators  

There are enhanced responsibilities for regulators and bodies operating in the employment and artificial intelligence field.   

Employment Tribunal awards and recommendations  

Many of the provisions in the Bill are backed up by the right to bring an Employment Tribunal claim which can result in a financial award and recommendations.    

Innovation provisions   

There are some innovation provisions to allow for regulatory sandboxes. This will provide for safe innovation and demonstrates that the measures in our Bill will not inhibit innovation.