Social Care & Youth Justice
19 February, 2026
In August 2026, more than 50 cities and counties across the United States had cancelled or deactivated Flock Safety cameras — the automatic licence plate readers installed across neighbourhoods, highways, and public spaces — following a nationwide backlash over privacy violations, civil liberties breaches, and evidence of racial profiling. People had been wrongfully detained. Protest attendees had been tracked. Law enforcement officers had misused the data to spy on former partners. The ACLU described it as the biggest grassroots privacy backlash in over twenty years.
But this is not just an American story. Closer to home, in February 2026 it emerged that the UK government’s AI-powered benefits assessment system was getting decisions wrong by a statistically significant amount, with evidence pointing to discrimination against marginalised communities. The Department for Work and Pensions had been using automated systems to detect fraud in Universal Credit claims — systems trained on historical data that reflects existing inequality. When AI is built on data that carries those patterns, and when it isn’t designed and scrutinised appropriately, the result can be that it reproduces or amplifies the very inequalities it was meant to overcome.
In UK policing, Amnesty International’s research into predictive policing systems has found serious concerns about violations of the right to a fair trial, the presumption of innocence, and the right to equality — with Black and racialised communities in deprived areas disproportionately affected. A National Physical Laboratory study found that facial recognition technology used in the police national database produces significantly more incorrect matches for Black and Asian people than for white people. As The Canary reported in February 2026, the AI lead responsible for a new £115 million police data centre admitted the system would produce discriminatory results.
These examples are not inevitable. They reflect choices — about institutional priorities, about who is involved in designing and scrutinising systems, and about what accountability looks like when things go wrong. Artificial intelligence is not a neutral tool, and how it is deployed depends enormously on the policy decisions and priorities that shape it.
“In 2026, inequality is coded. Across welfare offices, police departments, and financial institutions, automated decision-making systems now shape who receives public assistance, who is flagged as ‘high risk’, and who gains access to credit.” — Frost & Sullivan Institute, July 2026
Employment, housing, healthcare, education, policing, welfare — the systems that shape people’s life chances are increasingly governed by algorithmic decision-making. Often invisibly. Often without the knowledge or consent of the people affected.
The research is consistent: when AI systems are trained on historical data that reflects existing inequality, and when they are not designed and scrutinised with appropriate care, they can replicate and entrench those inequalities rather than correct them. A benefits algorithm trained on data from a system that has historically applied greater scrutiny to certain communities may continue doing so. A predictive policing system trained on decades of over-policing in particular postcodes may direct more officers to those postcodes. As the Parliamentary Office of Science and Technology sets out, this feedback loop is a known risk — and one that requires active design choices and ongoing oversight to address.
The policy and institutional context matters enormously here. How AI is deployed — who it surveils, who it supports, who it excludes — is not a purely technical question. It is shaped by institutional priorities, by procurement decisions, by who is consulted in the design process, and by the accountability frameworks that govern how these systems are used and challenged. As the Frost & Sullivan Institute sets out, the same technology can be used to widen access to services or to restrict it. To identify need or to flag suspicion. To include communities or to control them. Which version we get depends significantly on the choices made by those who commission, design, and deploy it — and on how much voice the people most affected have in those decisions.
There is a risk, in conversations about AI and social justice, of framing the issue entirely in terms of harm. And the harms are real and documented. But the full picture is more complex — and more hopeful — than that.
AI also creates genuine opportunities: for employment in new and growing sectors, for entrepreneurship, for social mobility, and for communities to use powerful tools to address challenges that matter to them. The question is not simply how to protect historically excluded communities from the risks of AI — it is equally how to ensure they have the skills, access, and opportunity to participate in and benefit from what AI makes possible.
Communities that have historically been excluded from technological change have often ended up on the wrong side of it — not because the technology was inherently hostile to them, but because they were not at the table when the decisions were made. The response to that is not just better regulation and scrutiny, important as those are. It is active investment in digital skills, meaningful involvement in design and governance, and genuine access to the economic opportunities that AI is creating.
That dual frame — risk and opportunity, protection and participation — is at the heart of why this conversation matters now. As Amnesty International UK has argued, the communities most affected by these systems need to understand them, have a genuine voice in the conversation, and be actively involved in shaping how they develop and are used. Not as an afterthought. From the start.
The UN’s Special Rapporteur on contemporary forms of racism has called out predictive policing for exacerbating the historical over-policing of communities along racial and ethnic lines. The European Parliament’s evidence on algorithmic discrimination has pointed to the Dutch child benefits scandal — where a racially biased fraud-detection algorithm contributed to the wrongful prosecution of tens of thousands of families and brought down a government — as a warning that the UK has not yet taken seriously enough. Written evidence submitted to Parliament has argued that generic equalities protections are not sufficient to prevent discrimination in AI systems, and that explicit statutory safeguards are needed urgently.
Yet in most of these conversations — policy consultations, tech conferences, government working groups — the voices of the communities most affected are largely absent. The people with lived experience of the benefits system, of stop and search, of school exclusion, of housing insecurity, are not at the table when the systems that govern their lives are being designed.
That has to change. And change does not happen by waiting to be invited.
For those working across youth work, social care, criminal justice, education, health, and EDI, this is not a distant policy concern. It is operational reality. The young person you are supporting may already have encountered or been affected by algorithmic systems — in school, in the benefits system, potentially in the criminal justice system. Understanding how those systems work, what they get wrong, and how to advocate within and against them is becoming a core practitioner competency.
But the practitioner role goes further than protection. It also includes helping the people we work with to see themselves as potential participants in, and beneficiaries of, the AI economy — not simply as people to be shielded from its risks. That means building digital confidence, challenging the assumption that AI is someone else’s domain, and supporting communities to develop the skills and agency to engage with these technologies on their own terms.
Our October webinar is designed for exactly that audience. Not a technical session for data scientists — but an accessible, honest conversation for frontline and grassroots professionals who want to understand what AI means for the communities they work with, and how to make sure those communities are part of the conversation about what comes next.
AI and Social Justice: Who Shapes the Future? — Free webinar, 7th October 2026, 4:00pm–5:00pm, online. Register free at ldhub.org.uk/webinars.
