AI privacy risks and ethical dilemmas

Artificial intelligence (AI) is a powerful technology that offers all kinds of opportunities for innovation, efficiency and improvement. But AI also brings privacy risks and threats, as well as numerous ethical dilemmas. This is especially true when it comes to processing personal data, spreading disinformation and influencing human behaviour. In this article we look at a few recent examples.

Versatile but double-edged chatbots

It is worth understanding that chatbots such as ChatGPT and Bing are extremely versatile. ChatGPT is a language model based on machine learning that can generate natural language from text input. Bing is a search engine that also has a chatbot function, which uses ChatGPT to hold conversations with users. Besides generating all kinds of text output, they can also do this for all kinds of software and programming languages.

Although ChatGPT and Bing are meant to provide useful information or hold interesting conversations, they can also give unexpected answers that do not match what the user intended or expected. Microsoft tried to integrate an AI language model into Bing that would be more powerful and more human than ChatGPT. Microsoft CEO Satya Nadella says it is important to develop AI that is “more aligned with human values”.

The Bing model seems to have more human traits than ChatGPT. But that does not mean Bing or ChatGPT are immune to errors, bias or misinformation. They can, for example, make statements that are controversial, unethical or even violent. For several users, including the test workers, this has led to confusion, fear or even shock and trauma.

More generally, many people are worried about the accuracy of the results, about bias and even about possible fabrications by the AI. Why do ChatGPT and Bing sometimes give unwanted answers? The reasons are not entirely clear. They may be influenced by the data they were trained on, which may be skewed or incomplete. Or they may be influenced by user input, when users have been provocative or misleading.

Risks and threats

AI is also used for all kinds of other purposes. The impact of fake news and disinformation, amplified by AI technology, is considerable. Smart algorithms can generate and spread large amounts of information in a short time, which increases the opportunities to manipulate opinions and sow division. This can lead to reputational damage or, at scale, even social unrest and political instability. That is why it matters to actively counter these forms of deception and to promote media literacy and critical thinking.

AI can also be used to synthesise or manipulate voices. With machine learning, hackers can imitate or alter voices to deceive or blackmail people. This can lead to identity fraud, reputational damage or even political instability. To limit this risk, regulation on data protection and privacy needs to come into force as soon as possible, and a person’s voice should be protected as, for example, “personal data”.

AI is also used to carry out or fend off cyberattacks. With ChatGPT and other AI, hackers can develop advanced malware or modify existing malware to avoid detection. That brings the risk of data breaches, system intrusions or ransomware attacks. Countering these threats is essential to keep data secure and prevent damage to systems.

AI is a fascinating and promising technology that can enrich and improve our lives. But AI also brings challenges and dangers that we must not ignore or underestimate. We need to be aware of the privacy risks and threats that come with using AI, and we need to act to prevent or limit them.

Further reading:

• AI-Powered Search Engines Driving Accuracy, Bias Concerns https://morningconsult.com/2023/02/22/ai-powered-search-concerns-bing-chatgpt/

• ChatGPT: Language Models are Few-Shot Learners https://arxiv.org/abs/2005.14165

• Voice cloning: a new privacy risk in the age of AI https://www.eitdigital.eu/newsroom/news/article/voice-cloning-a-new-privacy-risk-in-the-age-of-ai/

• How AI can help us fight cyberattacks https://www.forbes.com/sites/nvidia/2022/12/14/how-ai-can-help-us-fight-cyberattacks/?sh=6f9c0a7f6b8c

• How to fight fake news with artificial intelligence https://theconversation.com/how-to-fight-fake-news-with-artificial-intelligence-111318

Getting started

Want to use AI in a privacy-conscious way and in line with the rules? We can help with EU AI Act compliance consulting and broader AI consulting for organisations.