From reading MRI's to providing emotional support, Silicon Valley has made it abundantly clear what they see for our future, AI integration in every aspect of life, but as with any new development the public has made their concerns abundantly clear, the privacy, control, and ethics of AI have all come into question leaving many wondering, who's going to be held accountable when it goes wrong?
Guidelines around AI have existed long before the surge we're seeing now, with China being the first to implement its 'Three-year Guidance for Internet Plus AI Plan' in May of 2016, and many others following suit in later years. Despite early Strategies, it's only recently that we've seen true legal restraints being applied to AI development. America, the epicenter of this AI revolution, only just published its National AI legislative framework in March this year, a document that makes recommendations but carries no legal force of its own. Considering how much of the AI landscape America dominates, that is a remarkably late arrival to the table.
The most significant developments so far have emerged from Europe. The EU AI Act, the first comprehensive legal framework on AI worldwide, was introduced in August of 2024 and is expected to fully come into effect in August of 2027. It introduces a risk-based approach, categorising AI into four tiers. Unacceptable-risk systems, such as those that manipulate users or enable mass surveillance, are to be banned outright. High-risk systems, used in areas such as critical infrastructure, education, and law enforcement, are subject to strict compliance requirements, including human oversight and detailed documentation. It is, by any measure, the most serious attempt to draw a legal boundary around what AI can and cannot do.
Elsewhere, South Korea finalized its AI Framework Act in January 2025, strengthening transparency and safety requirements, while Japan enacted the AI Promotion Act in May 2025, pushing companies to cooperate with government safety measures, though without imposing penalties on those who do not.
Despite the push for regulation and the laws already in place, significant blind spots remain. Most glaringly in the question of liability, who deals with the consequences when AI makes the wrong call, gives the wrong information, or just straight up fails? You can examine a faulty break and determine where the design failed, but with AI, you can't do that. The innate nature of AI and its continual learning mean that hundreds of millions of people are partly responsible for whatever behaviour is outputted. Blame can be placed on companies and developers, but even if they follow every rule and regulation, there will still always be a chance of an AI behaving unpredictably or outside its programming, making it extremely difficult to evaluate any liability. The law just wasn't made with this kind of dynamic and self-modifying technology in mind.
Another major issue lies with the inconsistency of the policies surrounding AI. As of 2026, a minimum of 72 countries have proposed over 1000 different AI-related legislations, all with different standards and expectations regarding the development and use of AI. With no solid through across different countries' regulations, companies can take advantage of gaps, making 1000 different laws amount to very little.
So what should be done? The most sensible path forward seems to be one that incorporates clear liability chains, transparency, and international coordination. When it comes to liability, responsibility should be placed with those who are in the best position to stop the harm from happening in the first place. In this regard, developers and companies carry the heaviest burden, being responsible for limiting systems, predicting misuse, issuing warnings, etc, with failure to meet these safety requirements falling into already established negligence laws. Beyond liability rests transparency; regulators should require that any AI system making high-stakes decisions, whether that be medical, legal, financial, or otherwise, be auditable. If a company cannot explain why a system made a decision, it should not be permitted to use that system in situations that could cause drastic harm. In regard to international coordination, the disconnected laws surrounding AI across countries need to move towards something resembling a shared minimum standard, whether implemented at the state level or the federal level, laws should match a shared vision for the betterment of the majority.
AI doesn't seem like it's slowing down anytime soon, and the legal systems meant to govern it can't afford to either. We're at a critical point in the development of AI, and the decisions of those in power will decide what our future looks like. While we're seeing progress now, there's still much to be done if we want to minimize the potential harms of AI.
Sources & Further Reading
- "AI Model Can Read and Diagnose Brain MRI in Seconds." Michigan Medicine. michiganmedicine.org
- "AI Regulations Around the World — 2026." Mind Foundry. mindfoundry.ai
- "AI Tort Liability: Does Negligence Law Still Apply?" Justice Speakers Institute, 16 Dec. 2025. justicespeakersinstitute.com
- "Autonomous AI: Who Is Responsible When AI Acts Autonomously and Things Go Wrong?" Global Legal Insights, 10 Jun. 2025. globallegalinsights.com
- "The Case for AI Liability." AI Frontiers, 12 Jun. 2025. ai-frontiers.org
- "Emerging Legal Challenges: Artificial Intelligence and Product Liability." Product Law Perspective, 21 Oct. 2025. productlawperspective.com
- European Commission. "The AI Act." Shaping Europe's Digital Future. digital-strategy.ec.europa.eu
- "Global AI Law and Policy Tracker — Highlights and Takeaways." IAPP. iapp.org
- "Summary of Artificial Intelligence 2025 Legislation." National Conference of State Legislatures. ncsl.org
- "US AI Legislation Overview." Software Improvement Group. softwareimprovementgroup.com
- "Who's to Blame When AI Fails? The Struggle for Legal Accountability." CMU Public Law Review, 4 Sep. 2025. cmuplr.org