Artificial intelligence (AI) is advancing at an unstoppable pace, but its responsible integration into critical sectors such as healthcare and business depends on more than just technology. Governance has become the main obstacle , according to international organizations and recent studies. Both the WHO Regional Office for Europe and a Cloudera report indicate that the lack of robust oversight frameworks hinders the safe adoption of AI.
In the health sector, WHO/Europe presented the findings of its first Knowledge Community on Responsible AI, an online dialogue that brought together experts from 105 countries. In the business sector, Cloudera's The Great AI Re-Architecture study, based on data from more than 1.500 IT leaders, reveals that legacy data architectures limit AI scalability. Both reports agree that governance, not speed of implementation, is key to safe progress.
WHO Europe prioritizes governance in health
The WHO/Europe report, based on the first Knowledge Community on responsible AI in health, underscores that the main obstacle is not the technology itself, but rather the strength of the institutions responsible for regulating it. Participants from 105 countries identified fragmented and biased datasets, a lack of clarity regarding responsibilities, and gaps in AI literacy as persistent barriers. Enabling factors included robust governance frameworks , co-design with patients and healthcare professionals, transparency, and equitable access to data.
The document argues that ethics should be integrated into the daily functioning of healthcare systems, not treated as a one-off checklist. Continuous oversight is essential, including delaying or rejecting tools that cannot be responsibly governed.
Spanish companies: data reconfiguration and governance as a brake
A Cloudera study, conducted among 1.500 IT professionals and managers worldwide, shows that 84% of Spanish organizations believe their data architecture needs a thorough overhaul , significantly higher than the global average (72%). The implementation of AI has changed storage practices in 63% of Spanish companies, and 49% have migrated workloads from the public cloud to private or on-premises environments in the last year. Looking ahead to the next two years, 35% plan to increase their cloud spending, and 31% intend to adopt a hybrid model.
Governance and regulatory compliance have become the main obstacle: 49% of Spanish companies say that maintaining effective governance is more difficult with AI , and 32% cite security, governance, and compliance as reasons for changing their infrastructures. The most striking statistic is that 98% of Spanish organizations have delayed or canceled AI projects due to governance challenges. Sergio Gago, CTO of Cloudera, points out that traditional architectures were not designed for the flexibility and control that today's AI demands.
Tools for practical governance
Faced with these challenges, solutions are emerging that seek to integrate governance into the AI ​​lifecycle. Liferay has announced the public beta of Liferay AI Hub, a SaaS platform for creating and controlling autonomous AI agents , applying existing access and compliance policies. The solution is ISO/IEC 42001 certified and uses an open architecture compatible with models from Google, Anthropic, and OpenAI, through the standard MCP protocol. It includes low-code capabilities and multi-agent orchestration, with human oversight at key points.
The shift towards agent-based models requires prioritizing data traceability and sovereignty. The transition from pilots to production implementations will depend on combining algorithmic flexibility with strict access controls.
In short, AI governance has become a cross-cutting issue affecting health, business, and technological infrastructure. Reports and solutions presented in recent weeks indicate that institutional preparedness and oversight capacity are more important than the speed of adoption . For Spain and Europe, the challenge lies in building robust frameworks that enable innovation without losing control.
