Sam Altman: "Frontier AI progress should continue, but capabilities shouldn’t outrun safety, alignment, and monitoring."
Shared standards, independent audits, and safety checks earlier in development could be essential as AI systems become more powerful.
#AI#AISafety#OpenAI #FrontierAI #ArtificialIntelligence #AIGovernance
The world deserves confidence that American companies developing increasingly capable AI will act responsibly, especially as the trajectory of progress has steepened. Every frontier lab must deliver on this, and there is no reason any of us should come to work if we cannot.
We welcome a federal framework that sets consistent safety requirements for frontier AI. But we do not believe we need to wait for an anti-trust exemption or legislation to begin the work of providing this confidence. Consistent rules to manage frontier risk so that we can maximize the benefits are a good idea (and we are excited by ideas like independent auditors).
Years ago, companies like ours developed things like Responsible Scaling Policies and Preparedness Frameworks. Those were good for that moment, and focused primarily on the deployment of completed models, not what happens during their development process.
Today's shift to focusing on safe development and evaluation will need new tools. For example, at OpenAI we now formulate explicit safety cases in advance of frontier reinforcement learning runs we expect to significantly increase capability, in addition to the safety work we have long done in advance of model releases.
We hope that other companies will learn from our approaches and propose their own; we think shared standards for misalignment, monitoring, and safety will lead to better outcomes. We look forward to collaborating with our colleagues across the industry to formulate the best version of these.
When we talk about “pacing”, we do not mean “stopping”. Progress has been rapid and will continue to be. But it should be slower than it otherwise could be; interventions like safety cases and monitoring have significant costs.
Pacing will be well worth this cost; no amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring.
Where we will need the help of our government is for international coordination. But first we should do what we can ourselves.
We talked to Theo, Co-Founder and CEO of CloudNC, at IMTS!
@TeamCloudNC is using AI to help transform precision manufacturing by automating the difficult, repetitive, and time consuming programming tasks that happen behind the screen.
Together, Autodesk and CloudNC are partnering to bring AI-powered CAM programming to the industry, helping manufacturers spend less time on repetitive programming and more time making chips.
And if you have a great deep dish pizza recommendation, Theo is all ears.
Come see Autodesk and our partners at East Hall Booth #133310 at IMTS 2026!
This is Slack Surface, and this is a GAME CHANGER.
In a matter of minutes, talking to Slackbot, I had reporting with a customer user interface that required no development. Obviously, I had to make it a theme.
Natural language requests created INTERACTIVE modals with data from Salesforce.
Ill be doing more with surfaces for myself and my customers.
The possibilities here are only as limited as your imagination.
@Benioff@parkerharris@SlackHQ@SlackAPI@rbgavin
Your @Eve agents can now chat with users on iMessage 💬
Photon is now a first-party integration for @Vercel@Eve.
Get started with one command:
𝚎𝚟𝚎 𝚊𝚍𝚍 𝚌𝚑𝚊𝚗𝚗𝚎𝚕/𝚙𝚑𝚘𝚝𝚘𝚗-𝚒𝚖𝚎𝚜𝚜𝚊𝚐𝚎
Bring your agent to the conversations users are already having
Watch it in action ↓
smol machines release v1.16.x is here!
Introduces 1 major feature: incremental checkpoints.
Makes it possible to efficiently time travel through the history on your computer like you could with git.
Also makes branching and checkpointing 10x faster after the initial.
https://t.co/5OTF6fit7u
ATX builders: Matcha Morning is this Thursday. ☕
8:30 AM at Desnudo Coffee Downtown. Free matcha, no agenda, and a room full of people who'll happily look at your base.
Grab a spot: https://t.co/asdt5D2kWS
How do you build shared micromobility across multiple cities instead of one city at a time?
At Micromobility Europe 2026, Brock Rogers from Solent Transport shared how Southampton, Portsmouth, and the Isle of Wight are managed as one connected ecosystem with shared bikes and e-scooters.
The keynote covers everything from commercial viability and regional logistics to parking, accessibility and future legislation.
(00:00) Introduction
(01:42) UK's micromobility landscape
(03:48) Regional collaboration
(05:35) Moving to one operator
(07:45) Six million rides and user growth
(10:00) Urban vs rural operations
(12:20) Isle of Wight logistics
(14:18) Parking strategy
(16:35) Different parking models
(19:05) Accessibility and inclusion
(21:18) Carriageway parking
(23:40) Future parking expansion
(25:35) Safety data
(27:05) E-scooter legislation
(28:20) What's next
https://t.co/BZXs0DkqU4
Health insurance is overdue for a rebuild. @nara_benefits is taking it on from the ground up with AI, changing how people access and pay for care. Proud to back @sid_s17 and the Nara team.
Our 2026 AI Sentiment Report landed last week: original research with over 1,000 end users on how they feel about AI.
The results show that sentiment towards AI is rising. 49% of people said they've had a positive overall experience with AI, up 9 points from 2025. When we showed respondents an AI Agent resolving a real query mid-survey, that number jumped to 74%.
Trust in AI is also higher than it's ever been, but it's still conditional. People carry a lot of baggage from previous experiences with automation, so businesses need to earn their trust upfront.
Read the full report linked below to learn how.
I’ll be in London for RoboSim at @GoogleDeepMind this week for a sneak peak of some exciting things we’ve been working on at @SkildAI with @NVIDIARobotics. If anyone in attendance or nearby wants to talk open-source simulation or robotics more broadly, lmk!
The promise of AI for incidents isn’t finding signals, but connecting them.
A PR in one repo. An error in another service. A recent deploy.
Humans hunt for those relationships manually. Agents can follow the threads across systems and reason about how they fit together.
Almost 3,000 applications in, and some clear signals are emerging.
@LataPersson shares 5 areas she’s particularly excited by from the R[3]sidency × Construct applications.
24 hours left to apply ↓ https://t.co/gqmelZPvef
expo-device-hub, new from @expo, turns your browser into a local device dashboard 📱
- live stream of every iOS sim + Android emulator
- tap, swipe, scroll, type in the page
- boot and shut down without Xcode or Android Studio
- flip dark mode, and more
bun expo install expo-device-hub, then bun expo start. The link prints in your terminal :)
The best conversations from Summit week didn't happen on the main stage, they happened later that night.
We teamed up with Linear for the Lenny & Friends Afterparty at Anina. Tech themed cocktails, tacos, and a room full of product leaders.
Thanks to everyone who pulled up, and to Linear for co-hosting a great event!
Special thanks @danielyegarcia and @ciruyv for organizing!
A frontier model costs 133× more to win a single human preference than a cheap one.
Gemini 3.5 Flash Lite: $0.031 per win. Fable 5: $4.127 per win.
The gap in how often they win? Six points.
From iced coffees in Portland to sweaters in Melbourne, this season looks different across our global team, but our shared purpose remains the same ☀️❄️
👉 See what we have been up to this season: https://t.co/Ey7LB7hloc
Today we shipped DualEntry Tasks.
Think of it as a Kanban board for your ERP: work surfaces as it happens, agents handle more of the recurring stuff, and accountants step in when judgment or approval is needed.
The payoff is a close that keeps moving throughout the month, with LESS chasing, LESS clicking around, and MORE of the work already done by month-end.
More on what we built and why: https://t.co/icbDz7RgQs
See Tasks here 👉️ https://t.co/Y5D1SVyVMi
We're thrilled to share that Datadog has been named the Company to Beat for observability platforms in the August 2026 Gartner® AI Vendor Race research published on 19 August 2026. Learn more: https://t.co/ObfhD8aEdf