Welcome back! ChatGPT now lets you lock individual conversations behind a PIN, Face ID, or Touch ID on supported devices. Locked chats disappear from your regular history and get their own separate memory, although the feature doesn't change how OpenAI stores or uses the underlying data. Still, a welcome update for anyone who's ever handed someone their phone and immediately remembered the deeply questionable things they've asked ChatGPT.
3 New Claude Features You Can Actually Use at Work
Anthropic just rolled out some meaningful Claude upgrades. Three worth knowing about, with a prompt to try each:
Claude Motion: Upload a report and ask Claude to turn it into a short animated explainer instead of another 30-slide deck.
“Turn this into a 30-second animated explainer for our next team meeting, highlighting our biggest wins and key numbers.”
Type /motion to invoke. Team and Enterprise plans only. Exports to MP4.
Claude Dashboards: Connect Salesforce, Snowflake, or another data source and have Claude build a dashboard that stays live.
“Build me a dashboard showing this quarter’s revenue by region and how this week’s sales compare to last month.”
Available on all paid plans. Dashboards update as the underlying data changes.
Claude Haiku 5.5: A faster, cheaper model (roughly 75% less than the previous version) for high-volume tasks.
“Go through these 50 customer reviews, group them by common complaints, and tell me the top 5 things I should improve.”
Pick Haiku 5.5 from the model dropdown. Available on every plan, including Free.
Bonus! Claude Docs, Slides, and Design all moved out of beta the same day and are now on every plan (free included).
Which ones would you actually use? Hit reply and let me know.
— Matt
OpenAI Brings Its Models to Google Cloud
Via Reuters
OpenAI’s frontier AI models and Codex coding tools are coming to Google Cloud, expanding its reach to customers across all three major US cloud providers.
How it works: Google Cloud customers will be able to access OpenAI's models directly through Model Garden and purchase them through Google Cloud Marketplace. The integration also brings OpenAI models to Google's Antigravity development environment and Gemini Agent, with general availability expected in the coming weeks.
The enterprise advantage: Businesses can use OpenAI models without leaving the cloud environments where they already run applications and store data. Existing Google Cloud spending commitments can cover purchases, while built-in access controls and centralized data governance make adoption easier for organizations handling sensitive information.
An unusual partnership: The agreement puts OpenAI's models directly alongside Google's competing Gemini offerings. It also follows OpenAI's expansion to AWS earlier this year, while Microsoft remains its primary cloud partner. For Google, offering rival models could help keep customers on its platform even when they prefer someone else's AI.
The bigger picture: AI labs are realizing that winning enterprise customers isn't just about building the smartest model. It's about showing up wherever businesses already work. Rather than forcing companies into new ecosystems, labs are making their models easier to access and embed into existing workflows. The next phase of competition may be less about which model wins the benchmark and more about which one is easiest to put to work.
Anthropic Wants to Help Write AI Safety Rules
Anthropic is reportedly circulating draft legislation that would establish new safety requirements for frontier AI models, including circumstances that could block a model's release.
The proposal: The draft focuses on catastrophic risks involving biological weapons or cyberattacks. It includes scenarios where an AI model escapes developer controls or could foreseeably and imminently cause more than 50 deaths or over $1 billion in property damage. The thresholds are similar to those established in California's recently passed AI safety legislation.
How it would work: Independent evaluators registered with the Commerce Department would assess models for dangerous capabilities. If they identified an imminent catastrophic risk, they would have 72 hours to notify the department. The government would then need to seek a court order from a Washington, D.C., judge to halt the model's release. Federal agencies would not be allowed to block distribution on their own.
The unanswered questions: The proposal would allow AI labs to pay their own evaluators, provided compensation wasn't tied to the results. That raises questions about how genuinely independent those assessments would be. The legislation is also still a draft, with no guarantee Congress will take it up.
The bigger picture: AI labs aren't just waiting for regulators to write the rules anymore. They're increasingly trying to shape those rules themselves. Anthropic's proposal reflects a growing effort to establish concrete safety standards without handing regulators unrestricted power to slow development. The challenge is creating a system that protects the public while allowing innovation to continue, especially when the companies being regulated have a hand in designing the framework.
Generate and Edit Images With AI
Via KREA
Krea is an AI-powered creative platform for generating and editing images and videos. It offers access to more than 2,500 AI models, allowing creators to experiment with different styles, train custom models, and refine their work through a suite of visual editing tools.
How you can use it
Generate images and videos from prompts
Explore thousands of AI models and creative styles
Remove objects and edit images with inpainting
Train custom AI models for consistent visual styles
Generate variations and refine existing images
Pricing: Free and paid
Let AI Handle Your Missed Calls
Via Hola AI
Hola AI Voicemail is an AI-powered voicemail assistant that answers missed calls, speaks naturally with callers, and sends concise summaries by push notification, email, or SMS. It filters spam, organizes conversations into a searchable inbox, and works with your existing phone number through call forwarding.
How you can use it
Automatically answer missed calls
Receive instant summaries and transcripts
Filter spam and robocalls
Customize your assistant's greeting and personality
Search, organize, and replay past conversations
Pricing: Paid
Jobs, announcements, and big ideas
Railway discloses a security breach that exposed customer environment variables and API keys.
Anthropic launches Claude Duet, a proactive AI agent app for Pro and Max subscribers.
Google brings Gemini to Apple CarPlay, enabling hands-free Google Maps navigation in the US.
OpenAI deploys GPT-5.6 through Patch the Planet to fix critical PHP vulnerabilities worldwide.
ElevenLabs debuts Chimera, a new platform combining AI voice and video workflows.
Big labs make the headlines. But some of the coolest AI advancements fly under the radar. I walk through 11 breakthroughs you might have missed ⤵️
That’s a wrap! See you next week for more.
