Today’s artificial intelligence developments are connected by a question that is becoming increasingly difficult to avoid: who gets to shape AI, and who benefits from it?
UNESCO is attempting to turn ethical principles into practical skills, US policymakers are debating how AI-generated wealth can reach the wider population, publishers warn that AI could undermine the journalism needed to scrutinise existential risks, and the United Nations University is calling for governance models suited to countries with very different levels of digital readiness.
UNESCO and LG launch a global course on AI ethics
UNESCO and LG AI Research have launched a free massive open online course designed to help professionals translate ethical AI principles into practical decisions.
The Global MOOC on the Ethics of Artificial Intelligence is available through Coursera and is intended for technologists, researchers, policymakers, students and other professionals involved in developing, deploying or governing AI systems.
The course is based on UNESCO’s Recommendation on the Ethics of Artificial Intelligence, which was unanimously adopted by the organisation’s 193 member states in 2021. While that recommendation established an international framework, the new course focuses on applying its principles throughout the AI lifecycle.
Its ten modules cover fairness and inclusion, privacy and data governance, transparency, accountability, safety, security, environmental sustainability and global AI governance. It also incorporates case studies, practical frameworks and exercises examining the trade-offs that arise during real-world AI development.
The initiative reflects a broader shift from publishing responsible-AI principles to testing whether organisations can actually implement them. That requires more than compliance documents. Developers and decision-makers need the ability to identify risks before deployment, document their choices and demonstrate that high-impact systems meet meaningful standards for safety and fairness.
This move towards implementation is also visible in the emergence of independent behavioural-risk assessments for AI systems, which attempt to evaluate how models behave under realistic conditions rather than relying solely on technical benchmarks.
UNESCO’s course is primarily delivered in English, with translation options covering 11 languages. Its availability as a free global resource also addresses an important imbalance: access to AI-governance expertise remains concentrated in a relatively small number of companies, universities and technologically advanced countries.
Who should benefit from the wealth created by AI?
As investment pours into artificial intelligence, a parallel debate is developing around how the resulting productivity gains and corporate wealth should be distributed.
A CNBC analysis examines how the benefits of AI might be shared more broadly across the United States rather than accumulating primarily among technology companies, their investors and highly skilled workers.
The debate is becoming more urgent as AI systems automate tasks, reshape professional roles and increase the value of the companies controlling the most advanced models, computing infrastructure and datasets.
Proposals for distributing AI-generated wealth range from investment in education and workforce transition to employee ownership, public investment funds, tax-funded benefits and direct payments. Each option carries different implications for innovation, public control and the relationship between work and income.
Reskilling remains one of the least controversial approaches. If AI changes roles rather than simply eliminating them, employees need accessible routes into the new tasks and professions created around the technology. Previous industry research has warned that 92% of technology roles could be transformed by AI, strengthening the case for large-scale reskilling and upskilling programmes.
However, training alone may not address the complete economic challenge. Workers cannot all become AI engineers, and the number of new positions created may not perfectly match the occupations or communities affected by automation.
The deeper policy question is therefore one of ownership. If AI becomes a form of productive infrastructure, societies must decide whether its returns should flow exclusively to private shareholders or whether the public should receive a more direct stake in the value created.
AI’s economic success will ultimately be judged not only by productivity statistics or corporate valuations, but by whether it improves living standards across society.
AI could weaken the journalism needed to scrutinise AI
Artificial intelligence is not merely changing how news is produced. It may also be weakening the financial foundations of the institutions responsible for investigating its risks.
An analysis from the Bulletin of the Atomic Scientists warns that AI platforms can extract information from publishers, transform it into generated answers and retain audiences that might otherwise visit the original source.
This creates a fundamental contradiction. AI systems depend heavily on human-produced information, but their distribution models may reduce the revenue available to fund the journalists, researchers and subject-matter experts who create that information.
The problem is particularly serious for reporting on nuclear weapons, advanced AI and other existential risks. These subjects are technically complex, politically sensitive and often deliberately opaque. Meaningful coverage requires specialist knowledge, international sources and lengthy investigations—exactly the kind of journalism that cannot be sustained by producing large volumes of inexpensive content.
Research cited by the Bulletin found significant reliability concerns in AI-generated news responses. Some 45% contained at least one significant issue, rising to 81% when less severe problems were included.
Public trust also remains stronger when journalists retain control. Only 12% of surveyed respondents were comfortable with news created entirely by AI. That increased to 21% when a human was placed “in the loop,” 43% when journalism was human-led with AI assistance and 62% when it was produced entirely by a human journalist.
HIPTHER has previously examined how AI-powered search threatens publisher traffic and newsroom economics. As search platforms increasingly answer questions directly, publishers must rely more heavily on subscriptions, newsletters, events and direct audience relationships.
The issue extends beyond copyright or commercial negotiations. A society with fewer independent journalists will have less capacity to investigate the companies building AI, evaluate government claims or explain highly technical risks to the public.
If AI weakens the information ecosystem that holds it accountable, the technology may remove one of the safeguards required for its own responsible development.
Global AI governance must reflect unequal levels of readiness
The United Nations University Operating Unit on Policy-Driven Electronic Governance has contributed to international discussions on AI governance, sustainable development and public-sector deployment.
According to UNU-EGOV, its researchers presented work at the 39th annual meeting of the Academic Council on the United Nations System and the APEC Tech for Good Workshop.
One policy brief examined the potential of AI to support sustainable development in Small Island Developing States. Existing applications include disaster response in Fiji, renewable-energy planning in Palau and outbreak forecasting in Jamaica.
However, the research found that readiness remains deeply uneven. More than 70% of Small Island Developing States lack the quality data needed for AI, while many do not have a public AI strategy.
Closing this gap will require strong governance, open-source infrastructure, regional cooperation and long-term capacity building. Research into open-source technology and sovereign AI similarly suggests that countries need greater control over the infrastructure, models and data underpinning their AI systems.
UNU-EGOV also emphasised that responsible public-sector AI must be managed throughout its lifecycle. Governance cannot be treated as a final approval exercise after a system has already been designed. Problems introduced during data collection, development or procurement can influence everything that follows.
Human oversight should also reflect the risk level of each service. Citizens may accept significant automation for a routine process such as renewing a passport, while expecting meaningful human involvement in decisions affecting welfare, healthcare, policing or access to essential services.
This lifecycle approach complements efforts to create open and participatory models for governing AI, where accountability is incorporated into system design rather than added after deployment.
The bigger picture
Today’s developments show that responsible AI cannot be reduced to a single regulation, safety test or ethics statement.
It requires practical education for the people building AI, economic policies that distribute its benefits, sustainable funding for the journalists scrutinising it and governance frameworks that work beyond the world’s richest economies.
The common thread is participation. Ethical AI depends on who understands the technology, who owns its productive value, who has the resources to investigate it and who is represented when its rules are written.
Artificial intelligence may be developing at extraordinary speed, but its legitimacy will depend on whether institutions can ensure that progress remains accountable, inclusive and genuinely useful to the societies expected to adopt it.
