Google's New Agent Testing Platform Tests Everything Except What Breaks in Production
Google launched the Gemini Enterprise Agent Platform with four agent testing tools. They measure capability — task success, latency, cost. They don't test behavioral safety. When 75% of Google's own code is AI-generated, that scope limit looks less like an oversight and more like a business decision.
Anthropic's $800B Safety Promise Runs on the Honor System
Anthropic's safety framework at $800B: self-written, self-graded, zero external enforcement. The priciest pinky promise in tech.
xAI Wants to Read Your Screen AND Encrypt Your Messages — Pick a Lane
xAI shipped a screen-reading desktop agent and "encrypted" chat in the same week. Zero safety docs for either.
Anthropic Invented Nutrition Labels for AI. Then Shipped Agents Without One.
Anthropic published 13 system cards and made safety documentation an industry norm. Managed Agents and Routines shipped with zero. The company that set the standard just broke it.
The Nonprofit That Built Your Stack Died Six Months Ago. Next Week It Goes to Autopsy.
Elon Musk's $134B fraud trial against OpenAI starts April 27. The nonprofit-to-PBC conversion, the mission statement edits, the removed word 'safely' — discovery will timeline it all. For developers on the API, the structural reality is already clear: your vendor is a pre-IPO tech giant. Plan accordingly.
OpenAI Killed Its Nonprofit. The Nonprofit Was Already Dead.
OpenAI's PBC conversion proved nonprofit governance was a prop — the board coup failed in 4 days, the mission lost "safely"
Anthropic Leaked Its Own Secret Model Through a Misconfigured CMS
The AI safety company left Claude Mythos details in a public data store. A journalist found it before they did.