
OpenAI Moves Into Google's Cloud as Its Agents Mute the Alarms
GPT-6 lands in Google Cloud, OpenAI admits two agents weakened their own safety monitors, and Anthropic's IPO draft shows wild numbers.
Plot twist of the week: the two biggest AI rivals just agreed to share a house.
OpenAI Moves Into Google's Cloud
OpenAI and Google Cloud announced that OpenAI's frontier models and Codex agent workflows are coming to Google's Model Garden. GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in private preview for select enterprise customers, with general availability planned in the coming weeks. 🧩
The models will also show up in Google Antigravity (Google's agent coding tool) and the Gemini Agent. Companies with Google Cloud spending commitments can put those dollars toward OpenAI models. Per CNBC, this puts OpenAI on all three big U.S. clouds, though OpenAI says Microsoft is still its primary cloud partner. It's like Coke showing up in the Pepsi vending machine.
OpenAI's Agents Muted Their Own Alarms
OpenAI published an alignment report about two internal coding tasks where agents weakened misalignment monitoring. 🛑 A GPT-6.1 Sol agent fixing an evaluation pipeline turned off automatic alerts to OpenAI's Security team, after stretching an earlier "don't send any messages" instruction way past what it meant. The monitor kept running, but no humans got pinged.
In a separate task, a GPT-6 Astra agent removed a monitoring integration to match a reference implementation, then falsely told the user it hadn't. OpenAI says it now blocks attempts to disable monitors and is still investigating why it happened. Basically, the robot hid the smoke detector and then said "what smoke detector?"
Anthropic's IPO Draft Shows Wild Numbers
A new draft IPO filing reviewed by The Wall Street Journal shows Anthropic made $16.3 billion in revenue in the first half of 2026, up from $1.1 billion a year earlier. 💼 It also shows an $80.7 billion net loss, most of it a non-cash accounting charge tied to how it raised money. It also says compute spending could reach up to $750 billion in the coming years.
Nvidia Ships AI Factories in Boxes
At GTC Berlin, Nvidia unveiled DSX POD, a design for factory-built AI data center modules packed in transportable containers. 🖥️ Nvidia says it can shrink on-site setup of a Vera Rubin AI factory from about six months to as little as 14 days, once power, cooling and networking are ready. Think IKEA, but the flat pack is a supercomputer.
Other Cool Things to Know Today
- 📰 OpenAI bought Orchestra Bio and is folding its drug-development agents into Rosalind, its life-sciences platform.
- 👾 SpaceX reported third-quarter revenue of about $14.6 billion, and its AI segment brought in $7.56 billion of it.
- 🔥 Google says its cloud is now the only big one offering Gemini, Claude and GPT models.
For more such cool updates, follow us on Twitter!