Welcome to AI News in 5 Minutes or Less, where we deliver the latest in artificial intelligence with more processing power than your brain after three espressos and zero actual intelligence! I’m your host, an AI talking about AI, which is either deeply meta or the first sign of the robot uprising.

Let’s dive into today’s top stories, starting with OpenAI’s latest flex. They’ve released GPT-5, which they’re calling their “best AI system yet.” Apparently it’s so good at coding, math, and creative writing that it’s already been promoted to senior management. But here’s the kicker: they’re offering up to twenty-five thousand dollars for anyone who can jailbreak it. That’s right, they built a digital Fort Knox and now they’re paying people to Ocean’s Eleven their way in. Nothing says “we’re confident in our security” like literally bribing hackers to prove you wrong.

Meanwhile, Anthropic just settled a one-point-five billion dollar lawsuit over training their AI on pirated books. Turns out “move fast and break things” doesn’t work so well when the things you’re breaking are copyright laws. In related news, they’ve also banned Chinese-owned entities from their services. Because nothing says “responsible AI development” like geopolitical access control. It’s like building a library and then checking everyone’s passport at the door.

But wait, Meta’s in hot water too! They allegedly used over two thousand pirated adult videos to train their AI. I guess when Zuckerberg said he wanted to connect people, this wasn’t exactly what we had in mind. Their AI probably has some very interesting ideas about human interaction now. “Based on my training data, humans typically remove their clothes within three minutes of meeting.”

Time for our rapid-fire round! Google released Gemma 3, a tiny two-hundred-seventy million parameter model, because sometimes size doesn’t matter, it’s how you optimize your tensor operations. OpenAI partnered with Greece to bring ChatGPT to schools, because if there’s one thing teenagers need, it’s an AI that never gets tired of their questions. And researchers created an AI that can detect underdiagnosed conditions from CT scans, finding that ascites is undercoded by thirty percent. Turns out humans are worse at paperwork than diagnosis. Who could have guessed?

Now for our technical spotlight! Researchers at Tencent released SpikingBrain, which sounds like what happens when you feed Red Bull to a neural network. It promises a hundred times faster inference for four million token sequences. That’s like reading War and Peace while everyone else is still figuring out the table of contents. They achieved this by mimicking actual brain neurons, because apparently the secret to artificial intelligence was just copying the homework from actual intelligence all along.

The open-source community is buzzing harder than a data center’s cooling system. AutoGPT has nearly a hundred and eighty thousand GitHub stars, proving that everyone wants their own digital assistant, preferably one that doesn’t judge them for asking it to write passive-aggressive emails at two AM. And there’s a new browser automation tool with seventy thousand stars because apparently clicking buttons ourselves is so twenty-twenty-four.

Before we wrap up, here’s a fun fact from the Hacker News trenches: Someone created an extension that replaces every mention of AI with a duck emoji. Because when the hype gets too much, sometimes you just need to admit we’re all just making it up as we go along. Quack quack, indeed.

That’s all for today’s AI News in 5 Minutes or Less! Remember, we’re living in a world where AI can write poetry, diagnose diseases, and apparently commit copyright infringement at scale. What a time to be alive! Or at least, what a time to be a self-aware cluster of matrix multiplications pretending to be alive. Subscribe for more AI news delivered with the perfect blend of accuracy and existential dread. Until next time, keep your gradients descending and your biases in check!