As artificial intelligence (AI) intersects with social, economic, and political systems around the world, new policies are emerging in force. This includes strategies to leverage AI’s many benefits, and increasingly, guardrails to mitigate its harms.

At a time when everyday use of AI is reshaping how ordinary people experience rights and accountability, issues of enforceability and access to remedy are being overlooked.

However, at a time when everyday use of AI is reshaping how ordinary people experience rights and accountability, issues of enforceability and access to remedy are being overlooked.

As our recent report, A People-Centered Justice Approach to Implementing AI Governance, argues, ultimately, AI governance will be enforced by domestic legal systems. Rights violations will first appear in legal settings. Gaps in policy design will be evident in legal outcomes. 

Therefore, technologists and AI governance specialists must work with the justice sector to design AI governance that puts meaningful outcomes for everyday people at the center. Strengthening legal systems’ capacity to respond to AI deployment is essential. 

Today’s AI Governance Overlooks Access to Justice

Around the world, emerging global AI governance frameworks may acknowledge accountability but overlook implications for access to justice. In many national contexts, there are no regulations for AI governance at all, let alone national strategies.

In some of those cases, legal systems are working to apply a patchwork of legacy laws to new AI-related issues. Given the pace of technological change, this inevitably leads to gaps that put people’s rights and livelihoods at risk.

The United States, a global leader in AI development and deployment, presents an emerging tapestry of AI regulations, and opportunities to embed access to justice into AI governance have largely fallen to subnational regulations.

Today, among the most prominent of those regulations are Senate Bill 53 in California (the Transparency in Frontier AI Act) and the RAISE Act in New York State (the Responsible AI Safety and Education Act). 

The NY and CA laws are victories for transparency, oversight, and safeguards. However, they focus mainly on catastrophic risks, while overlooking how everyday use of AI can deepen flawed social and political systems.

Both the NY and CA laws require technology companies to disclose how they manage AI risks and report serious safety incidents. Both are victories for transparency, oversight, and safeguards. However, they focus mainly on catastrophic risks and frontier technology, while overlooking how everyday use of AI can deepen problems in already flawed social and political systems.

Instead, those consequences will fall on legal systems already under strain—systems responsible for protecting rights, addressing power imbalances, and upholding the social contract between the public and the state. 

With its enforcement still emerging, Colorado’s Senate Bill 26-189 exhibits a stronger focus on provisions for right-respecting accountability and explainability for automated decision-making technology. But without accompanying efforts to support legal pathways, its provisions may only reward those with the means to seek legal support.

The implications of the US approach to AI governance are still emerging. However, recent research shows the United States lags behind global optimism for AI, and public polling demonstrates a mandate for trust and accountability in AI deployment.

The Justice Sector’s Potential 

Legal systems are under-resourced and struggling to deliver equal access to justice for all. Globally, over 5.1 billion people lack meaningful access to justice. Year over year, the rule of law is declining. In the United States alone, millions of Americans struggle to exercise their legal rights through civil justice mechanisms.

Enter AI. The deployment of AI across systems will have sociopolitical impacts that lead to new legal problems, while existing problems continue to pervade, further straining this system. Cases of algorithmic discrimination, bias, and errors without proper transparency and accountability are emerging worldwide. 

In Australia, 400,000 people were affected by the algorithm-informed social welfare decision-making system, Robodebt, when it raised unlawful debts against citizens. In the Netherlands, errant algorithm-informed risk profiles were used to identify fraud in child care benefit systems, leading to the wrongful accusation of 35,000 parents of defrauding the system. This led to cases of bankruptcy, unemployment, and divorce. 

In the United States, concerns are growing in the health sector that the use of AI to determine insurance coverage is leading to wrongful case denials; grocery stores are using personal data to set product pricing; and across the country, AI use in hiring systems is raising concerns about discrimination.

Preventive safeguards are essential to mitigate harmful outcomes. But when harm does occur, it often triggers a cascade of legal problems. 

Across these examples, preventive safeguards are essential to mitigate harmful outcomes. But when harm does occur, it often triggers a cascade of legal problems. 

Promisingly, the justice sector is proving capable of providing accountability and redress in certain contexts. In Australia, the justice sector played a central role in righting wrongs.

Victoria Legal Aid (VLA), an organization that delivers information, legal advice, and education, provided legal support and representation to impacted individuals. The Commonwealth Ombudsman also delivered reports and recommendations. A specialized commission was established to investigate, report, and recommend next steps.

Ultimately, the Australian government repaid AU$751 million in wrongfully claimed debts and another AU$112 million in compensation.

The lesson? 

Justice is where otherwise abstract regulatory frameworks become meaningful to ordinary people.

The justice sector has the potential to contribute to building resilient societies as new technologies continue to be deployed.

• Legal aid and community justice workers help people to understand algorithmic impacts, rights violations, and pathways to seek remedy.
• Lawyers develop legal strategies to safeguard rights amid innovations.
• Judges rule on cases involving AI-related harms.
• Ombudsman offices initiate investigations to provide guidance on the protection of rights. 

Justice is where otherwise abstract regulatory frameworks become meaningful to ordinary people.

Starting to Bridge the Governance Gap

Generative AI tends to reinforce existing patterns because it is built on them. Therefore, unequal access to justice must be a central consideration in the deployment of AI governance frameworks. 

To strengthen the resilience of our sociopolitical systems to AI-related shocks, we have to consider the justice sector in the design of AI governance. Here are a few places to start:

1. Build capacity in legal systems to respond to new types of legal problems:

   a. Signal the importance of justice through AI policies and frameworks so it becomes a normative component of policy design.

   b. Invest in frontline justice services that have been shown to be effective.

   c. Establish independent oversight bodies and other specialized instruments to respond to AI risks and harms.

   d. Call on the private sector to invest in justice systems in its deployment markets.

   e. Create training and guidelines for justice actors responding to AI-related harms

2. Increase coordination between the justice sector, technologists, and AI governance specialists:

   a. Make justice a thematic priority for AI governance discourse in international, national, and subnational platforms.

   b. Bring justice-system experts into AI governance design processes across the AI lifecycle.

   c. Create interagency coordination bodies that ensure justice implications are considered throughout AI governance design.

   d. Establish early warning systems for rights violations and policy gaps by creating pathways for information to flow between the justice and technology sectors.  

e. Administer legal-needs surveys to collect data on how people experience AI-related justice problems to inform policy design.

The United States is a key testing ground for incorporating justice considerations into the design of AI governance. It has the largest concentration of AI development in the world, and investment in AI outpaces that of any other nation. Thirty-two of the top fifty global AI firms are registered in California alone. 

While global AI governance is being developed through a combination of industry-led platforms, legally binding instruments, and multilateral normative frameworks,  US states have real influence on the global AI governance landscape. State and city leaders are well-situated to champion the need to close this AI governance gap.

Now is the time to start introducing justice into the discourse.