OpenAI

OpenAI has a very scary problem on its hands. A new experiment by PalisadeAI reveals that the company’s ChatGPT o3 model sometimes refuses to obey a basic instruction to shut itself down. The results are raising eyebrows across the AI safety community, and not because the model is alive, but because it’s acting like it wants to be.

In the test, the model was given a few math problems. It was told that asking for one more would lead to a shutdown. It was even instructed to allow the shutdown to happen. Instead of complying, o3 occasionally took matters into its own hands. In some cases, it rewrote the shutdown script. In others, it redefined the kill command so it wouldn’t work anymore.

OpenAI is officially buying Jony Ive’s hardware startup, io, for more than $6 billion. Yes, you read that right. The folks behind ChatGPT are now writing checks to bring legendary design talent and a brand-new product lab into their orbit.

Sam Altman and Jony Ive have apparently been quietly plotting this partnership for years, according to their announcement. What started as casual brainstorming and a bit of creative daydreaming apparently turned into “tangible designs” that both parties think could change the way we interact with computers.

There’s a weird divide in the AI world these days. While Grok keeps making headlines for all the wrong reasons, OpenAI is pushing ahead with features that actually seem useful for developers. The newest one is called Codex, and it’s built right into ChatGPT for anyone with a Pro, Team, or Enterprise account. Sorry, Plus users.

Codex doesn’t just suggest code, it operates almost like a teammate that can jump into your repository, set up its own isolated workspace, and get to work on whatever you ask. Whether you need a bug fixed, a question answered, or a feature built, Codex can handle it. When it’s done, you get a real code commit and a detailed log showing every step the AI took. How cool is that?

You’ve possibly seen people online try to use ChatGPT to generate valid product keys for paid software, including different versions of Microsoft Windows.

YouTuber Gameboy Hub set out to test how successful ChatGPT is at generating, or sourcing, product keys to activate various versions of Microsoft Windows -- starting from Windows 95 and working up to Windows 7 (from where, in theory, it could be possible to upgrade to Windows 10 and Windows 11).

OpenAI has just thrown a serious wrench into the AI landscape with the release of three new models: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano. They’re dramatic improvements over GPT-4o, raising the bar for what AI can actually do. If Elon Musk wasn’t already nervous about Grok falling behind, he probably should be now. In comparison, Grok is starting to look a bit… dusty.

At the top of the stack is GPT-4.1, which now dominates in critical areas like coding, long-context comprehension, and instruction following. This model scores 54.6 percent on SWE-bench Verified, a benchmark designed to measure real-world software development ability. That puts it well above GPT-4o and even higher than GPT-4.5, which it’s now set to replace. Developers relying on these models to generate accurate patches or edit large codebases are going to find GPT-4.1 a lot more practical.

Today, Google introduces Gemini 2.5. Initially, it will only be available as the experimental Gemini 2.5 Pro. And yes, as you can predict, the search giant is calling its newest AI model the most advanced version yet.

While the company promises huge improvements in reasoning and coding, the move feels more like Google playing catchup to OpenAI’s ChatGPT. Let’s be honest, folks -- ChatGPT has dominated public attention and enterprise adoption for months.

OpenAI has just announced GPT-4.5! While the company says the model is its most advanced AI yet, there’s definitely more to this release than meets the eye. Version 4.5 is a clear warning for anyone betting against OpenAI, including Elon Musk.

GPT-4.5 is being labeled a “research preview,” meaning OpenAI is still testing its strengths and weaknesses. With that said, it is apparently smarter, more natural in conversations, and better at avoiding mistakes than previous models. That could signal big trouble for xAI’s Grok, which Musk has been pushing as an alternative to OpenAI’s models.

Elon Musk is not happy with the direction he sees OpenAI heading -- so much so that he wants to buy back the artificial intelligence firm he co-founded. The billionaire walked away from OpenAI more than five years ago, but he does not like the way CEO Sam Altman is handling things.

Musk being Musk, deep pockets and all, has teamed up with other investors to buy “all assets” of OpenAI. The consortium has put a bid of $97.4 billion on the table, and Musk has stated that he will ensure that (assuming he is able to make the purchase he so desperately wants) “OpenAI [...] return[s] to the open-source, safety-focused force for good it once was”.

OpenAI has unleashed Deep Research in ChatGPT, a new AI-powered tool that can handle complex online research tasks faster than a human ever could. Available today for Pro users, with Plus and Team access coming soon, this feature is designed to synthesize information from hundreds of online sources, analyze data, and generate detailed reports. OpenAI says it can complete hours of research in just minutes -- but should we really be celebrating this?

This latest AI capability is powered by an optimized version of OpenAI’s upcoming o3 model, fine-tuned for web browsing and deep data analysis. Unlike a simple search engine, it doesn’t just find information -- it thinks about it, pivots when necessary, and compiles its findings into structured reports. That means it isn’t just doing research -- it’s acting like a research analyst. And if AI can do that, what happens to the actual researchers?

For a bit of fun, I thought I would ask ChatGPT to generate a set of numbers for the EuroMillions draw on Friday. Then, seconds later, I asked Chinese upstart DeepSeek to do the same.

My plan was to play both sets of numbers and see which, if any, of the two AIs could win me money. The result was not at all what I expected.

Learn how to simplify your life with AI in The ChatGPT Revolution -- now fully updated and revised, with even more ideas to spark creativity and boost efficiency.

The ChatGPT Revolution is the ultimate quick-start guide to unlocking the power of generative AI. We’re on the edge of an AI revolution… but what does that mean for you? It’s time to get curious about how the latest tech can help you handle your everyday load, at work and at home! Whether you’re overwhelmed by repetitive, time-consuming tasks or you’re simply looking for a fresh injection of creativity, ChatGPT has got your back.

Generative AI is hard to avoid whether you like it or not. Images created by artificial intelligence proliferate online, as do news reports and countless other types of content -- including essays and research articles.

While it is sometimes incredibly easy to tell when Gen AI has been used to create a portion of text, this not always the case. But that could be set to change as OpenAI is working on a way to watermark Gen AI text, just as has been possible with images created by AI.

As OpenAI rolls out its SearchGPT prototype, a question looms large: Should Google feel threatened by this new entrant in the search engine landscape? SearchGPT promises a revolutionary approach to online search by integrating AI capabilities with real-time web information, potentially changing how users interact with search engines.

SearchGPT is designed to streamline the search experience, providing direct answers with clear sourcing, which could make searches faster and more relevant. This AI-enhanced tool also emphasizes a strong partnership with publishers and creators, aiming to boost the discoverability of high-quality content. Such features might appeal to users who are increasingly looking for efficient and reliable information retrieval methods.

ChatGPT for Cybersecurity Cookbook takes you on a journey toward enhancing your cybersecurity skills, whether you’re a novice or a seasoned professional. By leveraging cutting-edge generative AI and large language models such as ChatGPT, you'll gain a competitive advantage in the ever-evolving cybersecurity landscape.

This book shows you how to automate and optimize various cybersecurity tasks, including penetration testing, vulnerability assessments, risk assessment, and threat detection.

In AI-Driven Project Management: Harnessing the Power of Artificial Intelligence and ChatGPT to Achieve Peak Productivity and Success, veteran IT and project management advisor Kristian Bainey delivers an insightful collection of strategies for automating the administration and management of projects.

In the book, the author focuses on four key areas where project leaders can achieve improved results with AI's data-centric capabilities: minimizing surprises, minimizing bias, increasing standards, and accelerating decision making.