AI governance at a crossroads: Why corporate accountability must catch up with regulation

The global conversation on Artificial Intelligence (AI) governance has evolved rapidly over the past two years. Governments, regulators, international organisations and companies are racing to respond to the opportunities and risks created by increasingly powerful AI systems. Yet while regulatory activity has accelerated, a significant accountability gap remains between policy ambitions and corporate practice.

Around the world, governments are pursuing different approaches to AI governance. The European Union has adopted the AI Act, establishing a risk-based framework for AI systems. Other jurisdictions, including the United States, United Kingdom, Japan, Singapore and several emerging economies, have developed national strategies, voluntary codes or sector-specific approaches. At the international level, the UN has stepped up efforts to facilitate dialogue and cooperation through initiatives such as the Global Dialogue on AI Governance, recognising that AI impacts transcend national borders.

Despite this growing regulatory momentum, one challenge remains consistent across jurisdictions: ensuring that companies developing and deploying AI systems can demonstrate that they understand, assess and manage the risks their technologies create. This is where corporate accountability becomes critical. AI systems increasingly influence employment decisions, access to finance, healthcare outcomes, public discourse, content moderation, surveillance and many other areas that directly affect people's rights and wellbeing. Responsible governance therefore requires more than high-level principles. Companies must be able to show how they identify risks, engage affected stakeholders, implement safeguards and monitor impacts over time.

Recent findings from the World Benchmarking Alliance's (WBA) Digital Inclusion Benchmark underscore the scale of the challenge. WBA's assessment of 200 of the world's most influential digital technology companies found that none provided evidence of conducting comprehensive human rights impact assessments covering the AI systems they develop, procure or deploy. Even where companies reported undertaking assessments, disclosures were often limited, providing little information on scope, affected groups, AI functionalities reviewed or the outcomes of the assessment process.

These findings highlight a growing disconnect between corporate commitments and meaningful accountability. Many companies have published responsible AI principles and governance statements. Far fewer provide evidence that these commitments are being systematically implemented and monitored in practice.

This challenge is increasingly recognised by investors. Through WBA's Collective Impact Coalition (CIC) for Ethical AI, a growing group of investors representing more than USD 11 trillion in assets under management have engaged companies on issues including AI governance, transparency, human rights due diligence and risk management. Investors are seeking more than policy statements; they want evidence that companies understand and manage the societal impacts of AI.

The UN Global Dialogue on AI Governance provides an important opportunity to advance this conversation. While many existing discussions focus on technical safety, frontier models or regulatory frameworks, equal attention must be paid to accountability mechanisms that operate at the company level. This includes requirements for robust and transparent human rights impact assessments, meaningful human oversight, continuous monitoring of AI systems and effective stakeholder engagement.

The challenge is particularly acute because AI systems evolve rapidly. Governance cannot rely on one-off assessments or static compliance exercises. Instead, companies need ongoing processes to identify emerging risks, evaluate real-world impacts and adapt mitigation measures accordingly. Transparency around these processes is essential for regulators, investors, workers, users and affected communities.

At WBA, this understanding is also shaping the development of the Integrated Transition Assessment (ITA) framework. Within the new framework, responsible AI is primarily assessed through the social pillar because AI's most significant impacts are on people. Issues such as human rights, privacy, worker impacts, discrimination, stakeholder engagement and access to remedy are increasingly central to how companies manage technological transitions.

The future of AI governance will not be determined solely by regulation. It will depend equally on whether companies can demonstrate responsible behaviour, whether investors can hold them accountable, and whether global governance initiatives can establish common expectations across jurisdictions. As AI becomes embedded across the global economy, accountability, transparency and human rights due diligence must move from aspiration to implementation.