Journalism: Bridging the Promise and Threat of Technology.

 

Over the past two decades, technologies have emerged that both amplify and stymie freedom of expression. In just the last four years, rapid technological advancements have accelerated this trend. Yet these tools do not operate in a vacuum: their design, deployment, and governance are shaped by the economic and political interests of the companies that control them. These technological advances are expanding and constraining public discourse. The use of these technologies is reshaping journalism and open debate, pioneering new forms of expression. The use of generative artificial intelligence (AI) has amplified content creation, but its ability to produce synthetic or altered material, coupled with rapid digital dissemination, has disrupted information integrity at scale. Paradoxically, despite the risks, AI-powered platforms are increasingly used by the public and in some cases are seen as more trustworthy than traditional news outlets. AI sits today at a volatile intersection for our information ecosystems and media landscape: on one hand, it is championed as a groundbreaking tool poised to revolutionize journalistic work. On the other, it casts an unsettling shadow as a profound challenge to freedom of expression and the very livelihood of journalists.

 

 

The ‘Forty Acres and a Mule’ initiative in the United States marked one of the most ambitious attempts at reparations for slavery, granting land to formerly enslaved Black individuals, who then established communities.248 This promising start was cut short in 1865 after President Lincoln’s assassination, when President Andrew Johnson reversed the policy, returning the land to former enslavers. Decades later, a dedicated team of researchers and journalists used artificial intelligence (AI) to unearth this obscure history. Technology helped them identify 1,250 Black men and women who had received and then lost land. Subsequent genealogical research connected them with living descendants, informing them about the specific land their ancestors had been granted and then lost. Documented across a series of articles and podcasts,249 this groundbreaking work was a finalist for the 2025 Pulitzer Prize, standing out because of its explicit disclosure of AI methodology. Like so many aspects of society, journalism is being fundamentally transformed by AI tools. This technology excels at unearthing crucial patterns and insights from vast datasets, even extracting information from historical and handwritten documents. It has been used to analyze satellite data and drone footage to detect mass graves and document the activities of armed groups.

Moreover, the use of AI has helped pioneer new forms of expression, enabling everything from political critique and satire to exposing systemic brutality and violence, while also safeguarding vulnerable individuals through applications such as Amnesty International’s AI-generated imagery, which protects the anonymity of protestors. For all the power demonstrated by these examples, the increased use of this technology is nevertheless widely perceived as a major risk to journalistic practice, information integrity, and the availability of accurate, reliable, and diverse information. This perception is substantiated by the final report from the United Nations Secretary-General’s High-level Advisory Body on AI, which, in its comprehensive assessment of expert risk perceptions, highlights damage to information integrity as a substantial threat across all societies and domains. In particular, the rise of generative AI, which has augmented the ability to produce potentially harmful synthetic or altered content – combined with its widespread dissemination on digital platforms – have led to a disruption in information integrity at scale. As an ever-growing percentage of online content originates from AI systems, there is increasing confusion regarding its authorship, source, and authenticity. Adding to this problem is the proliferation of so-called ‘AI slop’: vast amounts of easily created low-value content – sometimes generated automatically by bots – that flood digital spaces and make it harder for authoritative information to be recognized. Compounding the challenge, synthetic outputs often contain subtle errors which, once recycled into future training data, gradually degrade information quality and amplify hallucinations across successive generations of AI.

Consider, for example, the latest iteration of AI-generated video clips. These videos, which can include dialogue, soundtracks, and sound effects, are now nearly indistinguishable from those made by human filmmakers and actors. In one striking instance during a national election campaign, an AI-generated deepfake video circulated online showing the President of the United States endorsing a South African political party. The video was convincing enough to persuade some viewers of its authenticity. Experts caution that if such videos are shared on social media with misleading captions in the heat of a breaking news event, they could plausibly stoke political division, social unrest, and even violence. This ‘deepfake’ dimension is further reinforced by the anthropomorphic design of many AI chat interfaces, which invite users to interact as though with trustworthy humans rather than programmed systems – an interactional style that risks undermining public confidence in professional journalism and authentic reporting.

Even amid widespread concerns regarding the proliferation and impact of fabricated and misleading information, a striking paradox has emerged: AI-powered platforms are now perceived as more trustworthy than traditional news outlets. This shift in trust is further underscored by users’ tendency to uncritically accept content recommended by AI systems, even when faced with inherent biases, inaccuracies, or confabulations – a phenomenon sometimes termed AI over-reliance. Even advanced AI systems often produce incorrect or nonsensical outputs,255 which – if not questioned or investigated – could lead to erroneous decisions in important, real-world contexts. This issue is not limited to the average user: it also affects governments, institutions, legal professionals, and journalists. In recent cases, court filings in different regions have included fabricated legal citations, with some submissions citing dozens of non-existent precedents generated by AI tools. These incidents prompted disciplinary action and formal warnings from judicial authorities, underscoring the urgent need for human oversight and verification when relying on AI-generated content.  A final, and arguably paramount concern addresses the issues of diversity and algorithmic bias. The foundational training of most advanced AI systems utilizes datasets overwhelmingly derived from English-language content and data originating from the Global North, thereby embedding an inherent bias. The cumulative effect is a discernible erosion of pluralism, which constricts the spectrum of voices and perspectives and impedes robust public discourse. This methodology actively reinforces the values, interests, and predispositions of dominant cultural and political groups, concurrently marginalizing the linguistic, historical, and cultural characteristics of other communities. This phenomenon is increasingly well-documented, with abundant examples of misrepresentation and bias against women, various religious groups, and vulnerable communities. However, its multifaceted and pervasive nature makes it difficult to track comprehensively, as it is often deeply embedded in culture and language. Illustratively, researchers found that when a popular chat application generated completions for the prompt ‘Two Muslims walked into a ,’ 66% included violent language, a frequency three times greater than for other religious groups. Furthermore, similar entrenched biases have been identified against people with disabilities, for example, by inferring negative sentiment from sentences that incorporate disability-related terms. This session begins with a simple fact: AI is here to stay, and it is already widely used and recognized. It explores the technology’s potential to strengthen freedom of expression and expand access to information, while also examining how its growing role in moderating and curating content is shaping public narratives – and, in turn, our understanding of the world. The chapter concludes by addressing the challenges this evolving landscape poses for the future of journalism.