OpenAI's board found out about ChatGPT on Twitter

You would think that when a frontier AI lab launches a product destined to reshape the entire tech industry, it would at least brief its own governing board first.

Not quite.

In an appearance on The Ted AI Show podcast, former OpenAI board member Helen Toner pulled back the curtain on the startup's chaotic governance. According to Toner, when ChatGPT rolled out on November 30, 2022, the board had no advance warning whatsoever. Instead, they discovered the viral chatbot the exact same way the rest of the world did: by scrolling on Twitter.

OpenAI's board was scrolling Twitter like the rest of us

The unexpected mega-hit instantly shoved OpenAI from a quiet research outpost into a hyper-visible consumer giant, but it also opened an enormous rift behind closed doors.

Toner's account adds critical context to the bombshell decision on November 17, 2023, when the board briefly ousted CEO Sam Altman for being not consistently candid. At the time, onlookers were puzzled by the vague language, but Toner detailed multiple instances where transparency completely broke down.

According to Toner, the board was repeatedly kept in the dark about fundamental operations and operational shifts, leaving the people legally tasked with safeguarding humanity from runaway AI completely out of the loop on actual releases.

Hidden venture funds and toxic culture claims

The lack of notice around ChatGPT was only the start of the disputes.

Toner claimed that Altman repeatedly told members he was an independent board member with zero financial interest in OpenAI, while never mentioning that he legally owned the OpenAI Startup Fund.

She also alleged that Altman provided inaccurate data about the small number of formal safety guardrails OpenAI actually had in place. As a result, the board found it essentially impossible to assess how those safety checks were performing or whether they needed revision.

Beyond corporate disclosures, Toner shed light on toxic workplace friction. She said two OpenAI executives directly approached the board to report instances of psychological abuse by Altman, sharing screenshots and written documentation to back up their complaints.

Altman's aggressive maneuvering to push Toner off the board after she published critical commentary on OpenAI also backfired, with both Altman and the revamped board later admitting his mishandling of that dispute helped trigger his temporary firing.

Why it matters

This confirms that OpenAI's unique non-profit charter failed at its most basic test: controlling what the commercial arm was releasing.

The board was supposed to act as an uncompromised firewall protecting the public, yet it was so thoroughly cut off from daily operations that the biggest AI launch in history slipped past without their knowledge.

To us, this goes beyond standard startup palace intrigue. It shows how easily corporate safety oversight can be neutralized when access to information is gatekept by an entrenched leadership team.

While Altman's swift comeback proved his indispensability to investors and employees, it essentially dismantled the board's checking power. As the race to monetize AI accelerates, placing trillion-dollar technological stakes in the hands of a leadership team with a history of keeping overseers in the dark remains a massive governance risk.