ChatGPT's new finance feature lets the AI chatbot take a look at any bank or similar accounts you care to open up for inspection. I was initially hesitant to try it out, but the tool only looks at the details of how you spend your money, and can't actually carry out transactions, so I agreed to let it analyze some of my accounts and offer its insights.

Finances is currently only available in the U.S. to Plus and Pro users on web, iOS, and Android. Setting everything up is as easy as using any other ChatGPT plug-in. You just select Finances in ChatGPT and then click Get Started then Connect with Plaid.

Finances uses Plaid to link to the accounts; you simply sign in and agree to let Plaid share the information. The service synchronized recent transactions and basically became an interactive, conversational database of my information. It's not that the AI can do things it couldn't before, but now you don't have to manually enter any finance details you want to ask it about.

ChatGPT looks at the receipts

At its suggestion, I asked ChatGPT, "How much did I spend eating out this year?" ChatGPT soon had a breakdown of my how much I had spent at various restaurants, including delivery, and broke down some of the patterns. It pointed out that it had gone up recently, which makes sense as the weather has became pleasant enough to walk into town more often to eat. It also identified a few spikes I could tell were around birthdays and family celebrations.

Next I asked, "What subscriptions am I paying for?"

That turned into one of the most practical moments of the experiment. ChatGPT gathered recurring charges across streaming services, software, and other monthly payments into one clear summary instead of leaving me hunting through statements. It also suggested reviewing whether every subscription still deserved a place in my budget.

"These subscriptions appear consistently each month, making them reliable recurring expenses.", it told me.

Financial chatter

Seeking a more detailed analysis, I asked ChatGPT, "What's changed in my spending this month?"

Instead of overwhelming me with every purchase, ChatGPT focused on the transactions that genuinely stood out. It noted a few large purchases at new locations, and asked if they were something I wanted more information on. I told it those were gifts, and the AI incorporated that fact into its response as not a new regular event but one that explained why overall spending had climbed this month.

"Those gift purchases are unusual compared with your recent history but they account for much of the increase, so your everyday spending has remained relatively consistent despite the higher overall total. Looking beyond one unusual month provides a more accurate picture of your longer term financial habits."

Using ChatGPT this way felt almost like the inverse of most financial software. Most budgeting apps expect you to learn their system first. ChatGPT let me ask questions however they occurred to me and then translated months of transactions into answers that were easy to understand.

That does not mean it replaces dedicated budgeting tools, and it certainly cannot make smarter financial decisions on your behalf. You still have to decide whether to cancel subscriptions, spend less on eating out, or save more each month. But I can see how it might make it easier for people who hesitate to look at their finances if they feel uncomfortable around spreadsheets. It didn't feel like math homework the way it often does.

Of course, it still relies on some trust in both OpenAI and Plaid, but as long as it's purely viewing and not actually touching the accounts, this could be a really useful, practical feature for ChatGPT users.

OpenAI's custom AI chip isn't just another attempt to loosen Nvidia's grip on AI hardware. It's the clearest sign yet that OpenAI is adopting the same vertically integrated strategy that transformed Apple over the past decade.

When OpenAI and Broadcom recently shared new details about Jalapeño, their custom inference processor, most of the discussion focused on Nvidia. Nvidia currently sits at the center of the AI industry, supplying the graphics processors that power everything from ChatGPT to image generators and coding assistants. Any attempt to reduce that dependence is naturally headline news.

For years, Apple has enjoyed a competitive advantage from making the most important parts of its products in-house. Instead of relying on someone else's processors or designing software around third-party hardware, it designed and built its own hardware and software. Competitors spent years trying to match that integration.

With its new custom inference processor, OpenAI appears to be building more than just an alternative chip. It's developing the same kind of vertically integrated ecosystem that helped transform Apple into one of the world's most valuable companies.

The chip is only part of the plan

When Apple introduced its M-series processors, the company aimed to build Macs that woke instantly and ran cool and quiet. Customers cared that everything simply felt smoother. OpenAI appears to be chasing a similar goal, even if the product is completely different.

Instead of laptops, it wants conversations that arrive faster. Building its own processor gives it another lever to pull that competitors relying entirely on third party hardware simply do not have.

Jalapeño is simply another piece of a much larger puzzle. The processor has been designed for inference rather than training. Training is the expensive process of creating an AI model as opposed to the inference done afterward. Every time someone asks ChatGPT a question, that's inference. Those billions of everyday interactions eventually become just as important as building the model itself because they determine both performance and operating costs.

Designing a processor specifically for those workloads gives OpenAI something that off-the-shelf hardware never fully can. It can begin tailoring the hardware around exactly how its own models think and respond, a more efficient method. And every improvement, whether in power consumption, speed, or networking, saves money and improves the AI experience.

OpenAI has been careful not to oversell the timeline, with broad deployment of the new chip still some way off. This is the beginning of a strategy rather than the final result.

Nvidia's challenge

Nvidia isn't going to panic right now, nor should it. Its processors still power much of today's AI boom. Demand continues to outstrip supply in many areas, and OpenAI itself remains one of its major customers. None of that changes because one new custom processor has appeared on the roadmap. What should catch Nvidia's attention is the pattern beyond OpenAI.

Google has spent years developing Tensor Processing Units. Amazon created Trainium and Inferentia. Microsoft has invested heavily in its own AI chips, as has Meta in custom accelerators for its expanding AI ambitions. OpenAI is now following the same path. Different companies have different technical goals, but they all seem to arrive at the same conclusion: as AI becomes a bigger part of their business, they don't want to depend entirely on someone else's hardware.

Of course, Apple designing its own processors certainly did not destroy Intel overnight. But there was a shift as Apple gained more control over pricing and product direction each time it replaced an external component with one of its own. The same could happen with AI.

Plus, OpenAI said its own AI models helped accelerate parts of the engineering process during chip development. AI is actually helping to make the hardware that will power its future iterations. That feedback loop may become increasingly important as chip design grows more complex. The future of AI may belong to the companies that own as much of the underlying machine as possible, regardless of where the models themselves rank.

If Apple's history is anything to go by, OpenAI is ready to be that company.

Anthropic has accused groups linked to Alibaba and its Qwen AI lab of carrying out a massive campaign to extract capabilities from Claude just by asking it a lot of questions, as first reported by Reuters. The AI developer wrote a letter to U.S. lawmakers alleging that Alibaba used nearly 25,000 fraudulent accounts to generate more than 28.8 million interactions and glean detailed, proprietary information about Claude.

Alibaba has not publicly responded to the allegations, and there has been no independent confirmation of Anthropic's claims, but simply leveling them has potentially enormous consequences. The sheer volume of accounts and interactions is eye-catching, but it's even more fascinating how it reveals a vulnerability in AI models that can give away their secrets.

AI developers may now have to worry that rivals can learn from those models without ever seeing the underlying code or training data through a technique known as model distillation. Essentially, AI models will inadvertently share deliberately obscured facts about themselves if a huge number of the right questions are asked. As an analogy, imagine taking a test about a book, but instead of reading the book, you ask the author one million questions about their life, their thinking, their experience writing the book, and several hundred thousand more questions. You'd probably have a pretty good chance of knowing everything they might have written without once cracking the covers.

Can you copy an AI just by talking to it?

Model distillation is a common technique used by AI companies to build variations of their models, especially smaller, faster options. But no company would be okay with a rival using their model to train the competition. But that's what Anthropic alleges. The fake accounts supposedly asked Claude a ton of very complex and detailed questions related to its advanced software engineering and agentic reasoning features. The responses filled in a picture of the model's workings, accelerating Alibaba's own development of competing AI systems, Anthropic claimed.

The conundrum is obvious. Large language models are designed to answer questions. Every answer teaches the user something about how the model behaves. You can't interact with an AI model, or a person, without giving up some information about yourself. Normally, that wouldn't matter, but at the scale Anthropic is claiming, conversations become reverse engineering.

It's not the first time Anthropic has alleged illicit model distillation. Anthropic levied similar claims against DeepSeek, Moonshot AI, and MiniMax earlier this year. And other companies, including OpenAI, have expressed concern that they have also been victims of the technique.

The glaring irony that the companies that used enormous collections of publicly available information, including licensed material, to train their AI models are now arguing about how those same models are valuable intellectual property is hard to ignore.

AI arms race

AI developers see their models' behavior as crucial to competing with rivals. If another company can reproduce much of that behavior by asking enough carefully designed questions, spending billions of dollars training frontier models starts to seem like a waste.

Anthropic claims model distillation can effectively transfer years of work on their part to another company for almost nothing. Anthropic asked lawmakers to take action and combat this problem as soon as possible. If leading models can be imitated so easily, there won't be much incentive to innovate, and the AI competition will only be about beating copycats. And picking the best models will be difficult, as a new AI model that matches an existing one's capabilities might be born of years of original research or simply copying an existing option.

Whether Anthropic ultimately proves its allegations, they have revealed that the next great AI battle may not be about building the smartest model. It may be about stopping somebody else from talking to your model and learning how it operates, one question at a time.

Woody the toy cowboy has faced a lot of frightening moments over the course of five Toy Story movies, but the actor voicing him for the last 30 years is trepidatious over the possibility that AI will supplant him. Tom Hanks told Entertainment Weekly in a new interview that the idea of AI replacing his voice is "a scary thought."

"Time is undefeated," Hanks said. "The question would be whether or not we could cobble together some version of me. Every word we have ever recorded in time in Toy Story is on digital media somewhere, so they could put together anything they would want."

Debates about AI in movies and film are raging even as the technology is deployed to reproduce the voices and complete performances of people who have passed away. Not to mention fully AI-generated films that otherwise employ human voice actors.

AI acting

Most entertainment projects with AI so far have generally involved the participation or approval of the performers themselves, but they also demonstrate how quickly the technology has matured. What once required painstaking effects work can increasingly be achieved with sophisticated machine learning models trained on years of existing material.

Hanks has himself been in AI-assisted productions, specifically in the film Here, which relied on AI to de-age him and co-star Robin Wright for part of the film rather than depending entirely on traditional visual effects.

Hanks' concerns extend beyond his own career. He has elsewhere suggested that his biggest worry is whether audiences will eventually stop caring whether a performance comes from a human being at all, not trusting that what they see isn't produced with AI. That feels like a more unsettling question than whether Woody could return without Hanks.

Future films

There's a little irony in Hanks' comments. The original Toy Story was a technological revolution that some saw as a threat to traditional hand-drawn animation. Three decades later, one of the stars who helped usher in that revolution is looking at the next wave of technology with understandable caution.

Caution doesn't mean dismissal of AI, of course. Hanks is known for being part of projects using advanced tools like motion capture, digital filmmaking, and experimental visual effects throughout his career. His concern seems more about what gets lost when technology starts replacing the people audiences are connected with in the first place.

And while each individual use of AI can be justified as another filmmaking tool, there's still a larger question about where AI assistance ends, and replacement begins. The answer will probably depend on what audiences decide they value rather than purely technical capabilities. Movie stars are more than collections of facial expressions and vocal recordings, and recreating the performance that audiences associate with Woody, that's more than just "good enough," would be a tall order.

ChatGPT's new Scheduled Tasks feature announcement this week — the ability for ChatGPT to now send reminders, handle recurring work, or monitor things — caught my attention immediately.

After all, AI assistants are all predicated on being reactive and needing your initial input. No matter how sophisticated the technology becomes, the responsibility usually remains firmly on your side of the screen. But if they are to truly be assistants, they have to be able to help you out when you might not remember to ask for their help.

Scheduled Tasks do require your input to begin with, of course. But you can't tell ChatGPT to remind you about something later. You can ask it to send recurring updates. You can keep an eye on a topic and notify you when something changes.

Unlike a lot of ChatGPT features, this one truly seems to give ChatGPT more of an actual assistant feel.

How to start using Scheduled Tasks

I started with a basic reminder of something I frequently forget. I asked ChatGPT to remind me to practice my saxophone three evenings each week.

Again, the setup was conversational. There were no complicated menus or automation builders. I simply described what I wanted. ChatGPT replied, "I'll remind you to practice saxophone three evenings per week. I'll keep the reminders encouraging and focused on making consistent progress."

Your Scheduled Tasks are all accessible from a new left-hand menu item, nestled between Projects and Apps, provided you are a ChatGPT Plus, Pro, Business, or Enterprise user. If you click the menu option then ChatGPT will suggest some things to try with Scheduled Tasks.

That evening I received the first reminder, with a couple of links suggesting songs to learn. Exactly the kind of thing that will help keep me on track.

If you've enabled notifications for the ChatGPT app on your phone, you'll get a notification that your scheduled task has completed. If you're using ChatGPT in a browser then turn on Desktop Notifications when it asks you. If email notifications are enabled for scheduled tasks in your ChatGPT settings, you'll also receive the result by email.

I then set up a more complicated evening reminder. I wanted a suggestion sent every day at 4 p.m. for a short but fun game to play outside with my child. The idea was to outsource a small piece of the fretting I sometimes feel to keep young kids entertained. Happily, not long before he returned home from daycare, ChatGPT suggested a fun and even slightly educational game of dinosaurs which he loved because of getting to run around and roar, but also helped him learn the names of several dinosaur species.

Getting a useful morning briefing

My third test was set for overnight. I told ChatGPT that I wanted a quick summary of local events or critical information to know every weekday morning at 8 a.m. Normally, this would involve checking several websites and social media. ChatGPT agreed to do so, and this morning I got a roundup of not only the day's weather, but details of yesterday's election news, the upcoming World Cup games, and how they might affect traffic.

Plenty of apps can recommend activities. What mattered was that the suggestion appeared before I had even started thinking about them or when it felt too late. Throughout the day, I kept noticing the same thing. Scheduled tasks were not saving enormous amounts of time. They were saving small moments of mental effort. Each individual task removed one tiny obligation from my internal to-do list.

That is the subtle trick scheduled tasks pull off so well. Instead of responding to a need, it anticipates one. Countless services can handle individual pieces of what happened during those two days. What felt different was having everything tied to the same conversation. That may be why the feature works better than you might expect. It remembers things so that you do not have to.