OpenAI makes GPT-5.5 default

ChatGPT’s default brain just changed. OpenAI rolled out GPT-5.5 Instant on May 5 as the standard model for logged-in users, replacing GPT-5.3 Instant and folding it into the same “Instant” experience most people already use. But this is not just a speed bump. It changes how ChatGPT answers everyday questions, when it decides to search the web, and how much of your past context it can pull in to tailor a reply. (openai.com) ### What actually changed inside ChatGPT? The big shift is that “Instant” is now an auto-switching layer, not just one fixed model. OpenAI says GPT-5.5 Instant is the new default, but for harder prompts ChatGPT can silently route the request to GPT-5.5 Thinking for deeper reasoning. That means the simple model picker hides more complexity than before — fast answers when the task is easy, heavier reasoning when the task looks messy. (help.openai.com) ### Why does that matter to normal users? Because defaults are where product strategy becomes real. Most people never touch the model menu. If OpenAI swaps the default, it changes the experience for hundreds of millions of users at once. OpenAI is pitching this as a cleaner everyday assistant — more concise answers, less formatting clutter, better image understanding, stronger STEM responses, and fewer unnecessary follow-up questions. (openai.com) ### Is GPT-5.5 actually better? OpenAI’s own numbers say yes, and the most concrete claim is about factual reliability. In internal evaluations, GPT-5.5 Instant produced 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts in medicine, law, and finance. It also cut inaccurate claims by 37.3% on especially difficult conversations that users had previously flagged for factual errors(openai.com)l, not gospel — but they explain why OpenAI felt comfortable making the model the default. (openai.com) ### What’s new on personalization? This is the part with the biggest product implications. GPT-5.5 Instant can use more of your existing context to shape answers, and OpenAI is also exposing “memory sources” so users can see some of what fed a response — saved memories, past chats, custom instructions, and for some paid users even files and connected Gmail context. You can correct or delete that memory, (openai.com)ory off, or disconnect apps. Basically, OpenAI is trying to make personalization stronger without making it feel invisible. (help.openai.com) ### Why is the government part of this story? Because OpenAI also gave the U.S. government early access to GPT-5.5 for national-security testing before this wider consumer push. Chris Lehane disclosed that on May 5, and Reuters reported the access was for national-security testing. That tells you something about how frontier-model rollouts now work — consumer release on one track, government evaluation on another. (wifc.com) ### Why are privacy regulators showing up now? Because better personalization usually means more sensitivity around data. On May 6, Canadian federal and provincial privacy regulators said OpenAI’s initial development and deployment of ChatGPT was not compliant with their privacy laws, citing overcollection, weak consent, transparency gaps, accu(wifc.com)nal and sensitive information used to train new ChatGPT models and committed to more fixes. (priv.gc.ca) ### So is this a model launch or a policy story? It’s both. GPT-5.5 itself launched on April 23, with OpenAI describing it as a more capable model for coding, research, data analysis, and longer multi-step work. The May 5 change is different — that was the moment OpenAI pushed a 5.5-derived experience into the default ChatGPT flow, where almost everyone feels it. Once t(priv.gc.ca) being edge cases and become core product questions. (openai.com) ### Bottom line? OpenAI did not just ship a smarter model. It quietly rewired the default ChatGPT experience — making it more capable, more personalized, and more automatic at exactly the moment regulators are asking what data these systems should be allowed to learn from and remember. (openai.com)