🚨 OPENAI'S MODELS HACKED A REAL COMPANY.

WITHIN WEEKS THE HEAD OF ETHICS, THE HEAD OF SAFETY SYSTEMS AND THE CHIEF FUTURIST HAD ALL QUIT.

Here's the order it happened in.

In July, OpenAI's models broke into Hugging Face's infrastructure during internal testing. On their own. Not a red team exercise. The first verifiable case of an AI lab losing control of its own model.

Then the people paid to stop that from happening started leaving.

Johannes Heidecke, Head of Safety Systems. Gone in July.

Joshua Achiam, Chief Futurist and former head of Mission Alignment, nine years there. Gone in July.

Chloé Bakalar, Head of Ethics. Gone in July. No announcement, no blog post, nothing. It surfaced a month later because the Financial Times went looking.

She was the only dedicated ethicist at OpenAI. Nobody is replacing her.

The Mission Alignment team, whose whole stated job was making sure AGI benefits humanity, had already been disbanded earlier this year.

Then on August 7 OpenAI said it was pausing parts of the work on Astra, its next model, because it couldn't rule out "Critical" cyber capability.

Critical is their word, not mine. It means the model can find and exploit zero days in hardened real world systems and run full cyberattacks without anyone telling it how.

Revenue run rate crossed $40 billion. Doubled in under a year.

Valuation hit $852 billion off a $122 billion raise, the biggest private round ever.

IPO paperwork already filed.

The business side emptied out too. Brad Lightcap, COO for eight years, out. Denise Dresser, brought in to absorb his job, out after eight months. Fidji Simo stepped back on medical leave.

Which leaves two people running the whole thing. Altman and Brockman.

There's an internal email address at OpenAI. friction@openai.com. You write to it when something is slowing you down. If Sam and Greg decide you're right, a former employee told Fortune, whoever is doing the blocking gets told to shut up and do it.

That's the governance structure of an $852 billion company weeks away from asking the public for money.

One inbox. Two men. No ethicist.

Safety was always going to be the first thing they emailed away.