OpenAI's whiplash week

Sam Altman had a rough week. After a spat over possibly imitating Scarlett Johansson’s voice turned into a broader criticism of the OpenAI CEO’s “beg forgiveness instead of asking permission” approach to doing business, the hits kept coming.

Repeatedly lying to the board was the reason a former OpenAI director gave in an interview for Altman’s brief firing last year. Then, a former OpenAI exec joined rival Anthropic, putting his comments — and those from other recent departures — about the company’s disregard for AI safety back in the spotlight. Even what should have been major news about training a new-and-improved AI model was more focused on a new AI safety team.

But on the other hand, OpenAI also accomplished more this week than any other in recent memory. It launched a specialized product to get its AI into more universities, enlisted consulting giant PwC as its biggest-ever enterprise customer, struck licensing deals with Vox and The Atlantic, and reportedly finalized a deal to provide AI on iPhones, which could be worth billions annually. The Information also pointed out that the executive departures meant Altman’s “primary foes” are no longer at OpenAI, cementing his control over its future.