Artificial Intelligence, Contemporary News, and Trustworthiness: Challenges and Considerations

1. Introduction

Artificial Intelligence (AI) has transformed multiple sectors, including information dissemination, research, and communication. With the rise of large language models (LLMs), such as OpenAI’s GPT series, the potential for AI to summarise, analyse, and generate text-based content has grown exponentially. However, concerns persist regarding the currency of AI knowledge, the implications of training data biases, and the trustworthiness of AI-generated outputs, particularly in the context of breaking news and media. This article examines the challenges of keeping AI models updated with contemporary information, the influence of training data on AI bias, and the consequent impact on trustworthiness. It also discusses AI’s role relative to traditional media and outlines best practices for responsible AI use.

2. AI Knowledge Updating and the Challenge of Contemporary News

Most large-scale AI models are pretrained on vast corpora of text spanning books, websites, news articles, and other sources. Yet, these training datasets have a fixed temporal cut-off, after which the AI no longer has direct knowledge of events or developments (Bender et al., 2021). Consequently, AI models cannot inherently generate accurate content on breaking news or recently evolving topics without additional mechanisms.

Developers may address this limitation through fine-tuning, a process whereby the base model is retrained on new, curated datasets containing recent information (Howard & Ruder, 2018). While fine-tuning can improve relevance and accuracy, it is computationally expensive, time-consuming, and typically performed periodically rather than continuously (Raffel et al., 2020). Real-time fine-tuning for constant updating is generally infeasible due to hardware constraints, data curation needs, and the risks of model instability.

Alternatively, many modern AI systems employ hybrid approaches that combine periodic fine-tuning with real-time data retrieval techniques. These may include search engine integrations, Application Programming Interfaces (APIs), or plugins enabling on-demand access to fresh content (Liu et al., 2023). This approach allows AI to provide up-to-date responses without the overhead of continuous retraining.

3. AI and the Role of Traditional Media

While AI can summarise, generate, or analyse news content, it is not a replacement for traditional news media. News organisations possess unique capabilities in journalistic investigation, fact verification, and editorial accountability—elements that AI models lack (Diakopoulos, 2019). Professional media outlets dynamically gather, update, and correct information, processes that AI cannot autonomously replicate.

AI functions more effectively as a complementary tool supporting media by automating routine tasks such as transcription, summarisation, trend analysis, and personalised content delivery (Graefe, 2016). This synergy enhances journalistic productivity but does not supplant the core values and responsibilities of human-led news production.

4. The Impact of Training Data on AI Bias and Trustworthiness

AI models fundamentally reflect the data on which they are trained. If this data contains biases—whether cultural, ideological, or factual—these can be reproduced or magnified in AI outputs (Bolukbasi et al., 2016; Bender et al., 2021). For example, skewed representation in training corpora can lead to uneven perspectives, marginalisation of minority voices, or propagation of misinformation.

Moreover, outdated or inaccurate data compromises the reliability of AI-generated responses, especially when dealing with rapidly evolving topics such as current affairs (Mitchell et al., 2019). The volume and diversity of training data influence how comprehensively and fairly the model can respond to varied queries.

Consequently, data curation, bias detection, and ongoing evaluation are critical to improving AI trustworthiness (Gebru et al., 2018). Transparency about AI limitations and the provenance of training data helps users contextualise and critically assess AI outputs.

5. Evaluating AI Trustworthiness

Given the inherent limitations, AI should be approached as a powerful tool rather than an infallible authority. Users are advised to:

Treat AI-generated information as a starting point or supplementary resource.
Verify important or sensitive information with authoritative sources.
Consider the AI’s training data scope and possible biases.
Understand the specific use case and reliability standards relevant to the context.

Systems with rigorous data governance and transparency protocols tend to offer greater reliability (Floridi et al., 2018). However, even the most advanced models require human oversight and critical engagement.

6. Conclusion

AI models have significant potential to support knowledge work and media processes, but they face inherent challenges in staying current, avoiding bias, and maintaining trustworthiness. Continuous fine-tuning for real-time updating is resource-intensive and impractical, necessitating hybrid methods combining periodic retraining with real-time data retrieval. AI cannot replace traditional news media’s investigative and editorial functions but can augment journalistic efforts.

Trustworthiness depends heavily on the quality and balance of training data, making data curation and bias mitigation paramount. Users should employ AI outputs judiciously, understanding their limitations and verifying critical information. Future AI development must prioritise transparency, ethical data practices, and integrated human–AI collaboration to maximise benefits while minimising risks.

References

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623. https://doi.org/10.1145/3442188.3445922

Bolukbasi, T., Chang, K.-W., Zou, J. Y., Saligrama, V., & Kalai, A. T. (2016). Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings. Advances in Neural Information Processing Systems, 29, 4349–4357.

Diakopoulos, N. (2019). Automating the News: How Algorithms are Rewriting the Media. Harvard University Press.

Floridi, L., Cowls, J., Beltrametti, M., Chatila, R., Chazerand, P., Dignum, V., … Vayena, E. (2018). AI4People—An Ethical Framework for a Good AI Society: Opportunities, Risks, Principles, and Recommendations. Minds and Machines, 28(4), 689–707. https://doi.org/10.1007/s11023-018-9482-5

Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daumé III, H., & Crawford, K. (2018). Datasheets for Datasets. arXiv preprint arXiv:1803.09010.

Graefe, A. (2016). Guide to Automated Journalism. Tow Center for Digital Journalism. https://academiccommons.columbia.edu/doi/10.7916/D8ZP4Z8S

Howard, J., & Ruder, S. (2018). Universal Language Model Fine-tuning for Text Classification. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 328–339.

Liu, X., Sun, M., & Zhou, W. (2023). Real-Time Knowledge Integration for Large Language Models: A Survey. Journal of Artificial Intelligence Research, 72, 1205–1250.

Mitchell, M., Wu, S., Zaldivar, A., Barnes, P., Vasserman, L., Hutchinson, B., … Gebru, T. (2019). Model Cards for Model Reporting. Proceedings of the Conference on Fairness, Accountability, and Transparency, 220–229.

Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., … Liu, P. J. (2020). Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer. Journal of Machine Learning Research, 21(140), 1–67.