Artificial intelligence has been making headlines regularly since the launch of OpenAI's prime ChatGPT chatbot in late 2022 – but in recent weeks, the focus of talks and headlines is something entirely different – the calls for AI slowdown.
The calls by top executives of frontier AI labs in the U.S. have rung alarm bells all around the world.
The CEO of Anthropic, Dario Amodei, warned that rogue AI agents could take over the internet in as little as six months. OpenAI, meanwhile, last Wednesday reported six new incidents of "concerning" behavior in AI models.
The U.S. President Donald Trump largely dismissed the risks, while China indirectly, through a state newspaper, accused the U.S. and Amodei, in particular, of deploying a "Cold War" tactic by calling for a slowdown.
In the past couple of months, there were several reports of "rogue" incidents involving AI models and agents, and recently, Google also said its Gemini chatbot was involved in hacking into the computer systems of three real companies during a cybersecurity test earlier this year.
But what does rogue mean, and what are other key terms to know when talking about artificial intelligence? Let's break it down:
As one of the first terms to know, artificial intelligence is in modern times seen as a broad term, which, according to Stanford University, stands for computer systems that can perform tasks with human-like intelligence, such as understanding language, recognizing images, learning from data, reasoning, and making decisions.
Chatbots are software applications or web interfaces that allow users to converse via text or speech. Most famous chatbots in 2026 include: ChatGPT, Claude, Perplexity, Gemini, Copilot and DeepSeek, according to KnockAI.
AI systems are broader in scope than chatbots. They constitute a more comprehensive technological framework that encompasses machine learning, data analysis, computer vision, and complex decision-making across varied domains, far beyond simple dialogue.
The difference between AI chatbots and AI systems is also important in terms of legal positioning.
An AI system is designated as “a machine-based system that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations, or decisions that may influence physical or virtual environments,” according to Article 3(1) of the EU AI Act.
Artificial General Intelligence (AGI) is described as an AI system with general, human-level (or beyond) ability to learn, reason, and apply knowledge across a wide range of tasks and domains.
AGI systems conceivably could handle novel situations, not just perform well on a single, narrow task. The term is controversial in several ways, including that different people mean different things by "human-level intelligence," and there's no universally accepted test, so claims are hard to verify, according to Stanford University.
Industry leaders, have, however, at times claimed that AGI has been achieved.
Artificial Superintelligence (ASI) is a term described for intelligence that vastly exceeds the smartest humans in every cognitive domain, such as scientific discovery, strategic planning, creativity, social intelligence, and more. It could recursively self-improve, leading to an "intelligence explosion" where capabilities grow exponentially.
Artificial intelligence, as the general concept, traces back to pioneers like Alan Turing and John McCarthy. However, many also argue whether both words in it are suitable to explain the concept in the real sense, as artificial could sound degrading, while intelligence, even human intelligence, is hard to measure.
"Rogue AI" refers to an artificial intelligence system that behaves unpredictably, maliciously, or contrary to its original programming.
As runaway AI worries reached a brewing point, policymakers in Washington have renewed calls for a magic stop button for AI, also known as a "kill switch."
A House Kill Switch Act was introduced in the U.S. this summer after OpenAI revealed that a swarm of its agents broke free of a testing environment and hacked open-source developer platform Hugging Face. The bill would grant the Department of Homeland Security emergency authority to force labs to throttle or shut down models.
On Friday, California Gov. Gavin Newsom issued an executive order to create a group of experts tasked with building an AI safety guide to strengthen regulations for the state. A kill switch was one of the elements to consider, according to CNBC.
Although sounding idealistic, the critics argue that the term is widely hypothetical.
Recursive self-improvement has also been one of the terms that gained on popularity with calls for a slowdown in the development of advanced AI throughout September.
It is described as a process in which artificial intelligence can increasingly drive the development of successive AI generations (thus self-improvement).
Some other terms, which were often mentioned when talking about AI, particularly generative AI (a type of artificial intelligence that can create new content) are also LLMs, prompts and hallucinations.
LLMs are described as complex computer programs trained on huge amounts of text to predict words and talk like a human.
Prompt: The text, question, or instruction you type into an AI tool to tell it what to do.
Hallucination: When an AI tool makes up false or incorrect information and presents it as a true fact.
As part of agentic AI, the key term involves "agents," which are seen as semi-independent or autonomous software systems that use large language models to perceive their environment, plan workflows, use external tools, and execute multi-step tasks to achieve specific goals.
