Artificial intelligence is usually discussed in terms of what it can do. It can help discover drugs, write software, analyse enormous datasets, translate languages and perform tasks that once required considerable amounts of human time and expertise. But there is another question that deserves greater attention: what is AI doing to the environment in which we decide what is true?
This was the question I explored recently in a paper prepared for the AI4Peace Global Symposium in Rome. My argument was that one of the most consequential challenges created by artificial intelligence isn’t simply the production of false information. It is the gradual degradation of the shared information environment on which functioning societies depend.
Generative AI has dramatically reduced the cost of producing convincing synthetic content. Images can be fabricated, voices cloned and plausible narratives created in seconds. The immediate response has understandably focused on detecting those creations, and better authentication, provenance standards and safeguards around synthetic media are all necessary.
But fabrication is only part of the problem. The more difficult question concerns what happens after information enters the digital ecosystem. Recommendation systems decide what billions of people see every day, and those systems have largely been optimised around engagement rather than accuracy, context or complexity.
Outrage, certainty and content that confirms existing beliefs often travel further than material that qualifies, complicates or challenges them. AI makes that dynamic more powerful by allowing content to be created at enormous scale, messages to be tailored for particular audiences and systems to become increasingly effective at capturing attention.
The danger isn’t simply that people will believe individual pieces of misinformation. It is that confidence in the information environment itself begins to weaken. When almost any image can be fabricated, any voice cloned and any event convincingly reconstructed, scepticism can spread from the false to the genuine. The eventual result may be less a society in which people believe everything than one in which they feel able to believe almost anything.
Democracies have never depended on everyone agreeing with each other. Disagreement is essential to them. But productive disagreement does require some common ground: an ability to establish that certain things happened and others did not, a belief that evidence matters, and institutions capable of establishing facts with a reasonable degree of credibility.
Once that common ground begins to disappear, political disagreement changes character. People no longer argue only about what should be done. They increasingly argue about what is real.
Recent events provide warnings of where this can lead. Following the murders in Southport in the United Kingdom in 2024, false information about the identity and background of the attacker spread rapidly online before basic facts had been established. During Romania’s 2024 presidential election, coordinated online activity involving social platforms, bot networks and manipulated content became part of a crisis serious enough for the country’s Constitutional Court to annul the first-round result. In the United States, voters in New Hampshire received AI-generated robocalls using a synthetic version of President Joe Biden’s voice encouraging them not to vote in the state’s 2024 presidential primary.
The technologies and circumstances were different, but the wider lesson is similar. The integrity of the information environment can no longer be treated as a peripheral technology issue. It is becoming part of the infrastructure of democratic society.
Governments have started to respond. Europe’s Digital Services Act and AI Act represent serious attempts to create greater accountability, transparency and safeguards around technology. They matter, but regulation will inevitably struggle with a problem developing this quickly.
Rules can require disclosure. Regulators can investigate platforms. Governments can prohibit particular practices. What legislation finds much harder to change is the incentive structure of an information economy built around capturing attention.
If outrage generates more engagement than nuance, there is a commercial reason to amplify outrage. If emotionally charged misinformation travels faster than a carefully qualified correction, removing it several hours later doesn’t erase its effect. Regulation can address harmful outputs without necessarily changing the systems that make those outputs commercially attractive in the first place.
That is why the response can’t belong to governments alone. Technology companies, AI developers, universities, researchers, civil society organisations and ethical institutions each hold different parts of the solution. Better technical standards are needed to establish where content comes from. Independent researchers need meaningful access to data so that recommendation systems can be studied properly. Education needs to place greater emphasis on critical engagement with information.
Technology companies also need to confront a broader question. Compliance matters, but so does the civic environment their design choices help to create.
None of this requires pessimism about artificial intelligence. AI is already accelerating scientific discovery, improving healthcare, increasing productivity, supporting environmental monitoring and helping people communicate across linguistic and cultural boundaries. Its capacity to improve human life remains enormous.
The same technologies that create new risks can also help address them. AI systems could improve the provenance and verification of information. Recommendation systems could be designed to expose users to credible alternative perspectives rather than endlessly reinforcing existing beliefs. AI could help identify coordinated manipulation, facilitate dialogue between communities and make high-quality information easier to find.
The central question is therefore not whether AI is good or bad. It is what objectives we choose when designing and deploying it.
This is closely connected to the thinking behind the Human Code at NJF Holdings. The framework starts from the principle that technological progress should ultimately strengthen human agency and human flourishing. That means asking not only what a technology can achieve, but what sort of society its incentives and systems encourage.
The next phase of AI will require more than increasingly capable models. It will also require companies, institutions and policymakers willing to think seriously about the information environment those models help create.
It is tempting to treat misinformation as somebody else’s responsibility. Governments should regulate it, platforms should moderate it and technology companies should detect it. All three have an important role, but the deeper challenge is also cultural.
Healthy societies depend on curiosity, intellectual humility and some willingness to engage with people who see the world differently. They need citizens who can distinguish evidence from assertion without assuming that disagreement itself demonstrates bad faith. Those qualities can’t simply be legislated into existence.
Technology can reinforce them or weaken them. The information systems being built today will shape how billions of people understand the world for decades, so they shouldn’t be judged only by how intelligent they become, how quickly they respond or how successfully they capture attention.
They should also be judged by whether they help people understand one another, preserve the distinction between authentic and synthetic information, and protect the common ground on which pluralistic societies depend.
AI doesn’t need to make us agree. But it should leave us capable of disagreeing about the same reality.
Nicole Junkermann is an international entrepreneur and investor and the founder of NJF Holdings. Through NJF Capital and the wider NJF Holdings investment platform, she has invested in technology businesses across artificial intelligence, cybersecurity, healthtech and other areas of emerging technology. Her Human Code framework explores how technological progress can strengthen human capability, resilience and agency.