# ChatGPT Adds PIN-Locked Chats—But Not Encryption | AI News Briefing Canonical URL: https://superpowerdaily.com/videos/db74b36f-fba0-45ad-8600-6f07273c3964 Published: 2026-10-21T08:00:25.000Z Episode date: 2026-10-21 Language: English Duration: 8:20 Publisher: Superpower Daily YouTube: https://www.youtube.com/watch?v=50_8pShh6pY ChatGPT is rolling out PIN-protected conversations that disappear from regular history and search, without adding end-to-end encryption or changing data controls; AI solver Darmok beat the human winner in the Times cryptic final. The Boston Symphony will premiere a work using an MIT-built system to connect live performances, while Eric Schmidt reportedly owns an Anthropic stake potentially worth billions—not a confirmed payout. ## Key takeaways - ChatGPT locks hide conversations locally without adding end‑ - Darmok combined improved models with a strategy of trying - Boston Symphony is testing AI as a live connecting partner - Schmidt’s reported Anthropic stake could have enormous pap ## Chapters - 0:38 Chat Privacy: https://superpowerdaily.com/videos/db74b36f-fba0-45ad-8600-6f07273c3964?t=38 - 2:13 Cryptic Reasoning: https://superpowerdaily.com/videos/db74b36f-fba0-45ad-8600-6f07273c3964?t=133 - 4:25 Live AI Symphony: https://superpowerdaily.com/videos/db74b36f-fba0-45ad-8600-6f07273c3964?t=265 - 6:08 Paper Wealth: https://superpowerdaily.com/videos/db74b36f-fba0-45ad-8600-6f07273c3964?t=368 ## Stories covered - OpenAI adds locks to ChatGPT conversations without changing data controls: https://superpowerdaily.com/posts/openai-adds-locks-to-chatgpt-conversations-without-changing-data-controls Selected chats move out of regular history and search. Their new messages use separate memory, but locking does not erase earlier memories or change model-improvement settings. - AI solver finishes Times cryptic final about three minutes faster than human winner: https://superpowerdaily.com/posts/ai-solver-finishes-times-cryptic-final-about-three-minutes-faster-than-human-winner Developer Sid Chatterjee credits stronger AI models and a revised solving process. A visual wordplay clue helped separate Darmok from several human finalists. - Boston Symphony will premiere a symphony that brings AI into live performance: https://superpowerdaily.com/posts/boston-symphony-will-premiere-a-symphony-that-brings-ai-into-live-performance Tod Machover’s “already and not yet” uses AI to connect musicians, not just generate sound. Its three movements explore democratic breakdown and renewal. - Eric Schmidt reportedly owns 1% of Anthropic, potentially worth up to $20 billion: https://superpowerdaily.com/posts/eric-schmidt-reportedly-owns-1-of-anthropic-potentially-worth-up-to-20-billion Bloomberg’s ownership estimate points to a major potential return on a 2021 investment. But the IPO valuation and whether Schmidt has sold shares remain unsettled. ## Transcript Automatically transcribed from the narration. [0:00] Welcome back to the superpower daily AI explainer for October 21st 2026 We've got a packed briefing today showing just how fast AI is moving on literally every front We're breaking down a new chat GPT privacy lock and AI Absolutely crushing a human cryptic crossword champion a live AI symphony over in Boston and Eric Schmidt's multi billion dollar Anthropic investment Let's get you up to speed We're hitting four quick chapters today First up chat GPT conversation locks [0:29] second AI wins a cryptic crossword third a live AI Symphony and finally Eric Schmidt's potential anthropic stake Let's jump right in Chapter one chat GPT conversation locks and privacy on shared devices So starting on October 20th open AI kicked off a gradual rollout of a highly requested feature for their free go Plus and pro users It's called locked chats Basically you can now protect specific chat GPT conversations behind a four digit pin or your device's authentication [1:00] you know like face ID Once a chat is locked It's completely hidden from your regular history your pin chats and even your search results Think about the practical use case here you hand your phone over to a friend so they can read a recipe But you've been brainstorming plans for their surprise party on chat GPT This feature talks all that sensitive stuff completely out of sight But we really need to distinguish between device level privacy and [1:23] back end reality here Locked chats do give you a separate memory space shared only within that locked section And yeah they hide the chat from local searches But and this is a big but they do not provide end to end encryption and they definitely don't erase anything You've already saved to your regular chat GPT memory So it essentially stops the person physically holding your phone from snooping But it does not change open AI's data retention policies [1:46] or your model improvement settings in the cloud There are also a few exclusions to keep in mind You can't lock temporary chats archive chats group chats shared links or anything with scheduled tasks or meeting recordings Oh And please take this warning seriously if you forget your pin and have to reset it you will permanently delete all of those protected chats in their files Literally zero recovery It's a strict local barrier Not a secure encrypted vault All [2:13] right moving from consumer privacy to complex reasoning Chapter 2 AI wins the cryptic crossword final where a system named our mock beat the human champion Earlier this October an AI solver named Darmok finished the times cryptic crossword final about three minutes faster than the human winner Liam Hughes the developer Sid Chatterjee out of UT Austin Powered our mock using Gemini 3.1 Pro and Claude Fable 5 But the real game changer was how he tweaked the solving [2:42] process instead of guessing answers early Darmok now waits it actually attempts every single clue on the board before committing a single answer to the grid It's mimicking human patients analyzing the whole puzzle before making a move Let me give you a concrete example of why this is such a massive breakthrough Check out this actual clue from the final wolf trap missing its lowest cross piece five letters Now this specific clue stumped three of the five human [3:10] finalists including the 14 time champ Mark Goodliffe Cryptic crosswords are wild because they require a totally different type of logic than your standard trivia crossword You're not just pulling a fact from memory You are deciphering a puzzle hidden inside a very strange sentence Let's actually walk through how Darmok nailed this visual wordplay Step one it read the word trap and found the synonym snare Step two it looked at the phrase missing its lowest cross piece It [3:38] literally figured out it needed to remove the lowest horizontal line from the letter E at the end of snare Step three without that bottom line the E becomes an F giving us the word snarf Which just like a wolf means to eat greedily I mean that is just incredible The AI isn't just regurgitating data It's following complex spatial visual instructions hidden in plain English Can an AI actually write a satisfying cryptic crossword Chatterjee says he still [4:07] can't find a clear pattern for the types of clues Darmok struggles with but he makes a great point Writing a cryptic crafting a sentence that reads smoothly in normal English but acts as deceptive instructions is way harder than solving one creativity is clearly still a tougher hurdle than decryption and Speaking of creativity Let's look at AI as an artistic collaborator Chapter 3 the BSO premieres a live AI symphony diving into Todd mack overs concept of the [4:35] already and not yet on October 22nd the Boston Symphony Orchestra is doing something wild Integrating a live AI system into a symphony premiere The piece is structured in three movements exploring democratic strain and renewal Movement one smash up actually has the string section actively crowded out by aggressive overwhelming brass and electronics Movement to wake up pivots into reflection flourishing with new textures and using the AI to spotlight individual players Then movement three stand up asks the [5:06] AI to help find common ground across the whole orchestra trying to harmonize individuality with the community Now while this MIT Media Lab AI is tasked with connecting the musicians human control is absolutely crucial here Todd mack over is the composer managing the studio electronics Manaswi Misra built the AI systems and Anna shown is doing the complex sound mixing live during the show So the electronic interaction heavily heavily depends on live human guidance It's designed to be [5:33] a true partnership not just handing the keys to a machine and letting it run wild on stage Which leads right into mock overs core question and the biggest limitation here Can current AI truly interpret expressive intent or is it just recognizing audio patterns Like is it genuinely feeling the music or just doing highly advanced math on sound waves Even mock over cautions that today's AI is much better at detecting audio features than actually understanding human emotion [6:01] This performance is a real time test in Symphony Hall to see if an AI can genuinely shape a musical exchange All right from the arts to some colossal financial stakes chapter 4 Schmitt's potential Anthropic stake and what looks like a massive return on paper So according to Bloomberg former Google chief Eric Schmitt owns about a 1% stake in the frontier AI company anthropic He picked this up back in 2021 by investing tens of millions during their [6:28] early series a funding round to put that into perspective 1% is roughly half the size of the estimated 2% stake held by anthropic CEO Dario Amadei The math on this gets pretty staggering when you look at the potential IPO valuations investors are currently throwing around If anthropic goes public at a hypothetical 1.5 trillion dollar valuation Schmitt's 1% stake is worth 15 billion dollars If it hits that 2 trillion dollar mark which is a number actively being [6:55] discussed that single percent jumps to a massive 20 billion dollars But before we get carried away there is a very crucial limitation We have to emphasize These are Hypothetical stake values entirely on paper not realized cash from a sale There's no set IPO price yet No confirmed date and honestly it's completely unclear if Schmitt has already sold off portions of his shares If he has that obviously completely changes his final payout Okay let's quickly wrap things [7:23] up for you today chapter 5 today's four key takeaways Let's distill the briefing Here is your cheat sheet for today's explainer first chat GPT added locked chats for local device level privacy But remember it is not end to end encryption Second the AI Dharmak beat a human crossword champ by combining better models with a really smart delayed grid entry strategy Third the Boston Symphony is testing AI as an active connecting partner in live performance rather than [7:51] just a pre programmed track And finally Eric Schmitt's early anthropic investment could be worth up to 20 billion dollars on paper But that's all pending the realities of a future IPO I want to leave you with this final provocative thought based on everything we looked at today Where is the line between AI as a tool we control and an independent collaborator We simply guide Read more at superpower daily com stay curious ## Editorial standards https://superpowerdaily.com/editorial-standards This episode reports information available at publication time. Consult the linked stories and their reporting sources for context and subsequent updates.