The emergence of artificial intelligence or AI technologies has transformed the existence of human beings. It has revolutionised the industries, reshaped societal operations and increased productivity both in personal and professional life. Alongside the advancements, there comes a crucial question with a legal dilemma: when AI causes harm or injures someone, who should be held responsible? The question is still unanswered as the traditional legal frameworks and enactments, particularly framed for human liability, cease to adapt to the complexities of AI.

While determining the criminal liability for AI-driven actions, one of the core issues arises as to the concept of intent. Fundamentally, the mens rea or the existence of a mental state such as intent or recklessness is pivotal in determining criminal liability alongside the criminal actions. However, in case of AI-driven omissions there is a lack of such consciousness or intention. Intrinsically, an AI system operates its actions based on algorithms and continuous learning process. This absence of human-like competence raises a crucial question regarding the determination of responsibility of AI systems: Shall AI be held accountable for its actions itself or it is the developers or operators or corporations using such AI tools that are to take such responsibilities?

As AI continues to advance, integrating it into the society requires balancing innovation with accountability to uphold justice and protect human rights. Legal systems must evolve thoughtfully and pragmatically, recognising both the transformative potential of AI and the ethical obligations it brings.

The question becomes further complicated as the AI systems encompass a distributed nature of responsibility. Most of the time, these systems involve multitudinous stakeholders such as manufacturers, programmers and end-users. Consequently, when harm occurs, such as in the case of automated vehicles, it becomes nearly impossible to determine who shall bear the ultimate responsibility. Is it the developer who is responsible for the injury or the end-user who did not update the system or the company that marketed the product who should be reliable?

Another concerning issue, in this case, is the opacity of AI working processes. Advanced AIs, especially those employing deep learning, often work as "Black Boxes" where the internal functions of such systems become difficult to interpret even by their creators. This lack of transparency makes it quite impossible to determine causation which is another fundamental factor within criminal law. Without understanding the decision-making process of AI including how it reaches to a conclusion and enacts the end-result derived from its functions, assigning blame becomes an uncertain and inherently complex process.

Additionally, the capacity of AI to learn and adapt things originates unpredictability. Unlike traditional machines, AI often works in a way that its developer never exclusively programmed or thought of. For example, Chatgpt-like AI systems that usually collect data from its users enriching its repository, at times may show results to be defamatory to someone or disseminate false information influencing the overall decision-making. These autonomous behaviors of AI systems raise questions regarding foreseeability and accountability.

To address these issues, legal frameworks should be developed to negotiate AI accountability. One potential approach is holding corporations accountable for the AI systems they deploy. This strategy emphasises the importance of thorough testing, transparency, and routine audits, pushing companies to focus on safety and proactively address potential risks. Some experts have even proposed the idea of granting AI systems a form of limited legal personhood, similar to corporations. This would enable them to assume certain responsibilities, such as facing fines or operational restrictions for their actions.

Regulatory sandboxes offer a valuable solution by allowing AI systems to be tested in controlled environments under legal oversight. These setups help regulators better understand AI's implications and fine-tune liability rules as needed. Additionally, hybrid models of shared responsibility are gaining support, where accountability is distributed among developers, operators, and users based on their specific roles. This approach encourages a culture of collective accountability.

Recent incidents highlight the challenges of assigning liability for AI-related issues. For example, the 2018 fatality of Elaine Herzberg, a 49-year-old woman, involving an Uber autonomous vehicle raised debates about the responsibilities of the safety driver, the company, and the vehicle's manufacturers. Similarly, courts are addressing cases like defamation caused by AI-generated content, holding platforms accountable for moderating harmful outputs. These situations reveal the pressing need for legal systems to tackle the unique issues AI presents.

As AI continues to advance, integrating it into the society requires balancing innovation with accountability to uphold justice and protect human rights. Legal systems must evolve thoughtfully and pragmatically, recognising both the transformative potential of AI and the ethical obligations it brings. By creating comprehensive and forward-looking frameworks, societies can fully harness AI's benefits while mitigating its risks.

The writer is student of law, Bangladesh University of Professionals and Vice President of Bangladesh Law Alliance.