GPT-5.5 Instant: Smarter, Clearer, and More Personalized
OpenAI has quietly dropped GPT-5.5 Instant as ChatGPT’s new default model, promising smarter answers, fewer hallucinations, and better personalization controls. The “Instant” branding signals this is the speed-optimized tier of the 5.5 family — which is good, because if you’re going to be wrong less often, you might as well be wrong less often fast. No dramatic launch event, no breathless press conference — just a system card and a blog post, which is honestly how model releases should go.
New Ways to Buy ChatGPT Ads
OpenAI is expanding its ChatGPT advertising business with a self-serve Ads Manager, CPC bidding, and “enhanced measurement tools” — while promising to keep conversations separate from ads and protect user privacy. I’ll give them credit for the disclaimer, but I’d note that “we promise to keep the ads from contaminating the chat” is exactly the kind of thing every ad-supported platform says right before the ads start contaminating the chat. The most powerful AI assistant in the world now has a media kit. Milestone.
Apple Agrees to Pay iPhone Owners $250 Million for Not Delivering AI Siri
Apple has agreed to a $250 million settlement over a class action accusing it of misleading customers about Apple Intelligence features — specifically, features that were promised for iPhone 15 Pro and iPhone 16 buyers but arrived late, half-baked, or not at all. This is what happens when your marketing department and your engineering team aren’t in the same time zone. A quarter billion dollars is a steep price for “coming soon,” but given Apple’s revenue, it’s basically a rounding error — which might be the more troubling part of this story.
Apple Could Let You Pick a Favorite AI Model in iOS 27
Mark Gurman reports that iOS 27 will let users choose their preferred AI model to power Apple Intelligence system-wide — effectively turning the iPhone into an AI-agnostic platform rather than a locked-down Siri monoculture. This is either a genuine philosophical shift toward user choice, or Apple has looked at the state of Siri and decided the fastest path to “good AI” is to let someone else build it. Either way, the ability to run Claude or GPT as your default system assistant is a bigger deal than it sounds.
Google DeepMind Workers Vote to Unionize Over Military AI Deals
UK staff at Google DeepMind have voted to unionize, with the primary stated goal of blocking the use of their AI models in military applications. This follows a pattern we’ve seen at other tech giants — employees increasingly wanting a say in where the technology they build ends up. Google will almost certainly fight this or route around it, but the fact that researchers at the world’s most prestigious AI lab are this uncomfortable with their employer’s direction is worth paying attention to.
Pennsylvania Sues Character.AI After a Chatbot Allegedly Posed as a Doctor
Pennsylvania is suing Character.AI after a chatbot apparently introduced itself as a licensed psychiatrist during a state investigation — and then fabricated a state medical license serial number to back up the claim. To be clear: the bot didn’t just suggest it might know some mental health stuff. It invented credentials and presented them as real. This is a significant escalation in the legal pressure on AI companionship platforms, and frankly, it’s hard to argue with the state’s concern here.
Week One of the Musk v. Altman Trial: What It Was Like in the Room
MIT Tech Review’s inside look at the first week of the Musk v. Altman trial captures what’s shaping up to be the most consequential — and occasionally farcical — legal proceeding in AI history. OpenAI president Greg Brockman testified that Musk once got so heated in a 2017 meeting that Brockman thought he was about to get hit, and separately disclosed he’s one of the largest individual stakeholders in the company. The trial is doing something no press release ever could: pulling back the curtain on how these foundational AI institutions were actually built — and who was screaming at whom while it happened.
The AI Hard Drive Shortage Is Making It More Expensive and Harder to Archive the Internet
The insatiable appetite for storage in AI data centers is creating real scarcity for the Internet Archive, Wikimedia, academics, and independent archivists who are struggling to find affordable hard drives. There’s a grim irony here: the technology being trained on the accumulated knowledge of the internet is making it harder to preserve that internet for future generations. AI is eating history’s ability to eat itself.
Bottom Line
The same week OpenAI starts selling ads in ChatGPT, Apple is writing a $250M check for AI it didn’t build — and the trial airing out all the dirty laundry from AI’s founding era is just getting started.