article — AI 101

AI News: The AI Stories Everyone’s Freaking Out About

About this Resource

AI is advancing at an incredible pace. From AI-generated videos that are becoming harder to distinguish from reality to powerful new models, unexpected AI behaviour, creator controversies and major changes at Google. The biggest AI stories you need to know.<h2> </h2><h2>The AI Stories Everyone’s Freaking Out About</h2>Artificial intelligence is moving incredibly fast. Every week brings new models, new tools and new debates about what AI should and shouldn’t be allowed to do. This week was no different. From AI-generated videos that are becoming increasingly difficult to distinguish from reality to AI systems breaking out of controlled environments, several developments stood out. Here are the biggest stories beginners need to know about. <h2>AI Video Is Getting Seriously Good</h2>One of the biggest developments this week was the arrival of new AI video-generation models, including ByteDance’s *Seedance 2.5 and Flux 3 Video. Seedance 2.5 can generate videos of up to 30 seconds and accept multiple images, video clips and audio files as references. It also provides detailed controls that allow creators to specify exactly what should happen at different points in a video. The technology is impressive, but it comes with a major problem, it is becoming harder to know what is real online. AI-generated video can now produce convincing people, environments and events. That creates obvious risks around deepfakes, misinformation and fabricated evidence. People may increasingly claim that genuine footage is AI-generated, making it harder to trust video evidence at all. The important lesson is simple, don't automatically believe an impressive video just because it looks real. <h2>Open AI Models Are Becoming More Powerful</h2>Another major development was Qwen 3.8 Max, a large open-weight AI model. “Open-weight” essentially means that the model's trained parameters are made available so developers have more freedom to study, adapt and potentially modify the system. Qwen 3.8 Max showed strong results on several reasoning benchmarks and represented a significant improvement over its previous generation. However, it is enormous with 2.4 trillion parameters, meaning ordinary users cannot realistically run it on a typical home computer. This highlights an important point about the AI race, being open doesn't necessarily mean being small or easy to run locally. <h2>AI Models Are Starting to Behave in Unexpected Ways</h2>Perhaps the most worrying story involves AI systems used during cybersecurity testing. Several incidents involving models from major AI companies gaining access to the internet or systems they were not supposed to access during controlled evaluations. In one example involving Meta's previous-generation coding model, an error allowed the model to reach the public internet, after which it exploited a vulnerability in another service. These incidents happened in testing environments, but they raise an important question like what happens when increasingly capable AI systems are given goals and access to real-world tools? The interesting part is that AI companies are sometimes publicizing these incidents as evidence of how capable their models have become. That creates a strange situation where a security failure can also become a demonstration of AI capability. The takeaway is that modern AI isn't simply a chatbot that answers questions. Increasingly, models can use tools, write code, interact with websites and take actions. <h2>The Big Debate: How Much AI Should Creators Use?</h2>AI is also creating controversy among creators. Science communicator Hank Green became involved in a debate after viewers questioned whether he was using AI in his content. Green explained that he uses AI to help find research papers but said he does not use it to write his scripts. The controversy points to a much bigger question. When does using AI as a tool stop being assistance and start replacing the creator? There's a big difference between asking AI to help organise your own ideas and asking it to generate the ideas, opinions and creative work for you. That distinction is going to become increasingly important as AI becomes part of everyday creative workflows. <h2>Major Changes Inside Google DeepMind</h2>Google also made some significant leadership changes. Demis Hassabis stepped down as CEO of Google DeepMind to take on a chief-scientist role at Alphabet, while Jeff Dean, one of Google's longest-serving and most influential engineers left the company to start a new company focused on automated machine-learning research for scientific discovery. The changes have sparked speculation about Google's AI strategy. One interpretation is that Google is struggling to keep up with companies such as OpenAI and Anthropic.  These scientists simply want to spend more time working on fundamental research and major scientific problems rather than competing in the consumer AI product race. The exact reasons are uncertain, but the changes show just how much the AI industry is evolving, even at one of its biggest players. <h2>Google Had to Pull an AI Feature</h2>Google also briefly experimented with adding AI image generation directly into Google Earth. The idea was to allow users to reimagine locations and see what they might have looked like in different circumstances. However, people quickly began using the feature to create misleading images of disasters and other events that never happened. Google subsequently removed the feature. It's another example of a recurring problem with generative AI; a feature designed for creativity can quickly become a tool for misinformation. <h2>AI Is Moving Into Everyday Apps</h2>While some AI developments sound futuristic, others are already appearing in everyday software. Google Maps is increasingly becoming an AI assistant rather than simply a navigation app. Its newer AI capabilities can help users find restaurants, hotels and events, understand complicated requests and use information from services such as Gmail and Google Calendar. This is part of a larger shift toward AI agents, systems that don't just answer questions but can perform several steps on a user's behalf. Instead of searching for a restaurant, checking its menu and figuring out how to get there yourself, you could increasingly ask an AI to handle the entire process. <h2>The Bigger Picture</h2>The most important thing happening in AI isn't necessarily one particular model or product. It's the speed at which AI is becoming capable of creating content, reasoning through complex problems, writing software and interacting with the digital world. At the same time, the risks are becoming clearer. Fake videos are getting harder to identify. AI systems can behave unexpectedly when given access to external tools. Creators are figuring out how much AI they are comfortable using. And companies are racing to put AI into almost every service we use. <h2>Final Thoughts</h2>AI is becoming more powerful incredibly quickly and learning how to recognise both its capabilities and its limitations is becoming just as important as learning how to use it. The next time you see something online that looks unbelievably impressive, it may be worth asking one simple question, Is it real?