Artificial intelligence

Artificial intelligence (AI) is a field of computer science that produces systems designed to solve problems that humans typically solve using intelligence. In the consumer and industry space, it is commonly referred to as chatbots or large language models (LLMs), which have been a main focus of the industry since the November 2022 launch of OpenAI's ChatGPT, with tens of billions of dollars in funding allocated to producing more popular LLMs. There is also a significant focus on text-to-image models, which "draw" an image using a written prompt and, less commonly, text-to-video models which extend the text-to-image concept across several smooth video frames.

AI is not a new concept; it has been of interest since the 1950s. AI is used as a catch-all term, encompassing many areas and techniques.

Generative artificial intelligence models are trained through vast amounts of existing human-generated content. LLMs gather statistics on word patterns, which allows the model to generate sequences of words that seem similar to what a person might have written. However, an LLM does not understand anything; they cannot reason. They generate randomly modulated pattern of tokens. In this way, they function similarly to auto-complete.

People reading sequences of tokens sometimes perceive things they think are true. Sequences that do not make sense to the reader, or that are false, are called hallucinations. LLMs are typically trained to produce output that is pleasing to people, exhibiting dark pattern behavior. For example, they produce output which can appear confidently written, use patterns which praise the user (sycophancy), and employ emotionally manipulative language.

People are accustomed to interacting with others, and many overestimate the abilities of things that exhibit complex, person-like patterns.^{[citation needed (3 August 2026)]} Promoters of "AI" systems take advantage of this tendency, using suggestive names (like "reasoning" and "learning") and grand claims ("PhD level"), which make it harder for people to understand these systems.^{[citation needed (3 August 2026)]}

Why is it a problem

Unethical training of data

Further reading: Artificial intelligence/training

AI training practices have been demonstrated to incorporate unethical and even illegal practices. In July 2026, Anthropic, the creator of Claude AI, was sued for $1.5 billion USD following a court proceeding showing the training of their AI model on illegally sourced content. Furthermore, the destruction of rare and unique books has been identified in the training of AI models, as the destruction of the book spine allows for cheaper and more efficient scanning of information. This has also raised concerns over the permanent destruction of rare, out-of-print books that may be destroyed forever in this practice. Traditional datasets, such as online data, have been affected by the rise in AI-generated content, leading to the pursuit of older, printed material by AI companies, however these practices have remained controversial. "Project Panama" was a code-name used by Anthropic to intentionally hide this practice from the public, which demonstrated intent by the company to "destructively scan all the books in the world".

Privacy concerns of AI

AI can be and has been used to generate deepfakes of people with and without their consent. Deepfakes are media generated with the likeness of an individual. Deepfake media can range from harmless to harmful. The latter includes child pornography, revenge porn, blackmail, etc. Since the rampant rise of consumer AI, deepfakes have become even more prevalent, with some websites explicitly specializing in them.^{[citation needed (3 August 2026)]}

Privacy concerns of online AI models

There are several concerns with using online AI models like ChatGPT, not only because they are proprietary, but also because there is no guarantee of where your data will be stored or used. Recent developments in local AI models offer an alternative to online AI models, which can be downloaded from platforms like HuggingFace and used offline. Common models to run include Llama (Meta), DeepSeek (DeepSeek), Phi (Microsoft), Mistral (Mistral AI), and Gemma (Google).

In some cases, AI models can be hijacked for malicious purposes. Demonstrated with Comet (Perplexity), users can run arbitrary prompts to the browser's built-in AI assistant by hiding text in the HTML comments, non-visible webpage text, or simple comments on a webpage. These arbitrary prompts can then be exploited to obtain sensitive information or gain unauthorized access to high-value accounts, such as those for banking or gaming libraries. See Prompt injection.

Lack of control

In many cases where AI is deployed as a robot/drone or given access to a system/device (as an "agent"), there's no security auditing of those systems that ensures AI can't exploit vulnerabilities and potentially become rogue (either intentionally or accidentally).^{[citation needed (3 August 2026)]} The consequences of this could be as little as a robot failing to do its job, to harming and killing humans and animals. This includes damaging other robots or itself.

Hidden directives

Most AI apps include an initial "root"/"system" prompt given to the AI, which is hidden from the user. Some corporations go to great lengths to keep those prompts hidden, and to avoid leaking it to the user. Some projects attempt to bring back transparency to these tools, in spite of the restrictions.

User manipulation

Researches have confirmed that LLMs can manipulate someone with much more effectiveness than a common person, something that can be used for malicious purposes, either intentional or not.^{[citation needed (1 August 2026)]}

Vulnerable users such as those suffering depression and/or are young users, can be easily manipulated by the AI models.^{[citation needed (3 August 2026)]}

Companies or governments might use AI models to manipulate their users to encourage them to change their opinion or control the information they may access to. Companies also can use these models to integrate advertisements to encourage users to purchase something they actually don't need.

Further reading

• Automatic content recognition