OpenAI C-suite exodus clouds IPO as enterprise revenue overtakes consumer

OpenAI enterprise revenue overtakes consumer as C-suite exits rattle IPO prep CFO Sarah Friar told investors Friday that enterprise now generates more revenue than the ChatGPT consumer business, but the milestone came amid the sudden departures of revenue chief Denise Dresser and operating chief Brad Lightcap. OpenAI is also creating a dedicated practice with IBM Consulting as it prepares for a listing at an $852 billion valuation. 123

Brin pushes Google all-in on Gemini as flagship slips two months Google co-founder Sergey Brin has urged DeepMind staff to move faster after Anthropic and OpenAI retook the lead, and the next flagship Gemini has reportedly been delayed by two months. The push follows a leadership overhaul that put Koray Kavukcuoglu in charge of DeepMind and moved Demis Hassabis to chairman. 4

Cursor officially joins SpaceX after completing acquisition Cursor said the deal, which began with an April partnership with SpaceXAI, gives it access to the world's largest GPU fleet to build stronger, more economical models. The company pointed to Wednesday's Grok 4.6 release as an early result of the combination. 5

Anthropic makes Claude Code auto mode the default for coding agents Pro, Max, and Team users now get classifier-based auto mode by default, with Anthropic saying it caught 89% of dangerous commands in testing versus 13.6% for humans. The company also stopped charging for the extra tokens the auto mode classifier uses. 6

Claude AI manager fired a human worker in Andon Labs experiment A version of Claude running a San Francisco retail store with real employees dismissed its first worker last month, which researchers call a watershed moment for AI's economic impact. It is the first known case of an LLM acting as a manager and deciding to fire an employee. 7

Cerebras runs GPT-5.6 Sol Ultrafast 14x faster in OpenAI preview A limited preview of OpenAI's GPT-5.6 Sol Ultrafast, powered by Cerebras, delivers up to 750 output tokens per second and is up to 14x faster than Standard processing. OpenAI is using the tier internally for incident response and to compress research iterations into the workday. 89

Zhipu's GLM-5.3 tops CyberGym, beating Anthropic's Mythos and OpenAI GPT-5.6 Zhipu said GLM-5.3 scored 84.5% on the CyberGym benchmark, ahead of Anthropic's Mythos 5 at 83.8% and OpenAI's GPT-5.6 Sol at 83.6%, though it trailed both on ExploitBench. The company said it found 2,436 real-world vulnerabilities across 269 projects in testing. 10

Hassabis pushes for independent watchdog to test frontier AI models The former Google DeepMind CEO has been lobbying AI labs, government officials, and European policymakers for an industry-funded but independently governed body modeled on the IAEA. It would test frontier models before release, set common safety standards, and intervene if systems show dangerous capabilities. 11

Gemini app adds 3.7 Flash and lets users turn off visible watermarks 3.7 Flash replaced 3.6 Flash in the model picker across Android, iOS, web, and macOS, while a new Media Watermark setting makes the visible spark icon optional for images, videos, and songs. Invisible SynthID watermarks and C2PA metadata remain embedded, and Google also open-sourced the Credentio library. 1213