Why governance keeps falling behind AI's rapid advances
The core problem is a fundamental mismatch in speed and nature. AI models, especially generative and autonomous systems, can be updated in weeks or even days, while legal and regulatory frameworks take years to draft, debate, and implement. A 2025 study [1] explains that AI's autonomy, algorithmic opacity (meaning even developers often can't fully explain how a model reaches a decision), transnational reach, and unprecedented scale and speed erode state control over rule-making and overwhelm regulatory institutions. This isn't a minor lag — it's a structural breakdown where the very tools of governance (laws, courts, agencies) were designed for a slower, more transparent world.
The problem extends into the courtroom. A 2026 analysis [4] of digital and AI-derived evidence in criminal justice found that existing legal standards — specifically Rules 901 and 702 under the Daubert framework — are ill-equipped to handle probabilistic evidence from proprietary AI systems. The authors note that these rules were built for deterministic, physical evidence (like a fingerprint), not for the probabilistic outputs of machine learning models. This mismatch creates constitutional and ethical risks, especially when such evidence is used in sentencing and corrections. In short, the legal system's tools for evaluating evidence are already outdated for the AI era.
What might work: adaptive, participatory governance models
Despite the bleak diagnosis, the same research points to concrete solutions. The 2025 study [1] advocates for a reconfiguration of legal frameworks toward adaptive, participatory, and ethically grounded governance models — systems that can update more fluidly, involve a wider range of stakeholders (not just regulators and companies), and embed ethical checks from the start. This is not a vague wish; the study bridges legal theory and political science to offer a theoretical and policy-oriented framework for reasserting the role of law in regulating AI-driven decision-making.
Similarly, the 2026 evidence-review study [4] proposes a structured integration framework that ensures scientific validity, transparency, contestability, ethical safeguards, and institutional accountability. The key idea is that governance must be built into the technology's lifecycle, not bolted on after deployment. Both studies [1][4] converge on the same conclusion from different angles — one from legal theory, one from courtroom practice — which strengthens the case that adaptive, multi-stakeholder, and transparency-focused frameworks are the most promising path forward.
The global and private-power dimension: who really governs AI?
Governance isn't just about speed — it's also about power. A 2022 study [2] on China's role in global AI governance reveals that the geopolitical landscape limits even major state actors' ability to shape norms. China, despite its ambition to become an AI superpower by 2030 and its desire to shift from a 'norm-taker' to a 'norm-shaper,' faces enormous challenges in demonstrating leadership in nascent global AI governance regimes. This means that even when nations try to set rules, geopolitical tensions can block consensus, leaving a governance vacuum.
Meanwhile, private companies are filling that vacuum. The 2025 study [1] explicitly warns that AI's characteristics shift governance power to private AI actors — the companies that develop and deploy the models — thereby diminishing democratic accountability. A 2021 study [3] on how manufacturing firms scale AI capabilities shows that companies focus on agile customer co-creation, data-driven delivery operations, and scalable ecosystem integration — all driven by business logic, not public oversight. When governance is left to private actors, the priorities are efficiency and profit, not democratic deliberation or equity. This private-power shift is a key reason why traditional governance frameworks struggle: they are trying to regulate entities that move faster, know more, and operate across borders.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2021 to 2026, 2 from 2024 or later, 2 in Q1 journals, collectively cited 808 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 71 papers retrieved from a database of over 500 million.
Sources used in this answer
AI and the Rule and Role of Law: Reshaping Legal Regulatory Frameworks to Address Emerging Challenges
Argues that AI's autonomy, opacity, transnational reach, and speed erode state control and shift power to private actors, advocating for adaptive, participatory, and ethically grounded governance models [2025 conceptual/legal analysis].
Shaping AI’s Future? China in Global AI Governance
Finds that China faces enormous challenges in realizing its ambition to lead global AI governance due to geopolitical constraints, limiting its ability to shift from norm-taker to norm-shaper [2022 policy analysis].
How AI capabilities enable business model innovation: Scaling AI through co-evolutionary processes and feedback loops
Identifies three critical AI capabilities (data pipeline, algorithm development, AI democratization) and shows that scaling AI requires business model innovation focused on agile customer co-creation, data-driven delivery, and ecosystem integration [2021 case study of six manufacturers].
Moving Faster Than the Rules
Demonstrates that existing legal evidence standards (Rules 901 and 702 under Daubert) are inadequate for probabilistic AI evidence, proposing a structured integration framework for transparency, contestability, and accountability [2026 legal analysis].
CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities
Presents a dataset of 1,445 writing sessions with GPT-3 to reveal its context-dependent language, ideation, and collaboration capabilities, arguing that large interaction datasets enable more incisive examination of AI capabilities [2022 HCI study with 63 writers].