OpenAI's Real Mission, in Its Own Words

OpenAI told the IRS last November that its mission is to ensure that artificial general intelligence benefits all of humanity. The word "safely" does not appear.

That single line from the nonprofit's 2024 IRS 990 filing IRS Filing is what Elon Musk's lawyers have been building toward for years. It is the document both sides have been waiting to put in front of a jury. Jury selection begins Monday in Oakland federal court.

Musk is seeking damages calculated by his expert at approximately $134 billion — traced from roughly $38 million in early donations that his side says have compounded into wrongful gains now approaching that figure Fortune Ars Technica. Current filings cite figures up to $150 billion; the range reflects a specific disgorgement calculation versus a broader damages request. Two claims survive: breach of charitable trust and unjust enrichment Fortune.

The contrast with earlier filings is not subtle. In 2022, OpenAI described its purpose as building AI that safely benefits humanity. The word sat in the mission line for years. The 2024 version has no qualifier The Conversation Fortune.

The nonprofit that holds the OpenAI mission now sits at the center of an $852 billion enterprise — a valuation that would make the expected IPO, if it happens, the biggest in technology history SF Standard. The nonprofit parent holds 26% of OpenAI. Microsoft holds 27% The Conversation. Per the 990, the nonprofit reported $691,000 in revenue and $7.78 million in expenses for fiscal year 2024. Total assets: $14.7 million ProPublica Nonprofit Explorer. The entity worth hundreds of billions is not the nonprofit. The nonprofit is the governance wrapper around the entity worth hundreds of billions.

This is the structure Musk's lawyers are asking a jury to find breached a charitable trust. The two surviving claims survived after the judge dismissed fraud claims at Musk's own request. Twenty-four of twenty-six original claims are gone. What remains is narrower and, for OpenAI, potentially more dangerous: not fraud but governance.

The witnesses scheduled to testify include Microsoft CEO Satya Nadella, former OpenAI CTO Mira Murati, cofounder Ilya Sutskever, and board member Shivon Zilis Implicator.ai. Nadella's testimony could be significant given that Microsoft holds 27% and is separately on the hook for $13.3 to $25.06 billion in disgorgement if Musk prevails Implicator.ai. The liability phase is expected to run through mid-May; a remedies phase before the same judge without a jury is scheduled to begin May 18 Implicator.ai.

The 990 also shows that OpenAI formed a Safety and Security Committee in May 2024, chaired by director Zico Kolter, to oversee critical safety and security measures related to model development and deployment IRS 990 filing via The Conversation. The committee was stood up in the same year the mission statement was revised. Whether that represents genuine institutional concern about AI risk or governance theater is exactly what the trial is supposed to determine.

What the IRS filing does not contain is a clear accounting of how the nonprofit's 26% stake translates into actual control. The 990 describes the nonprofit as the governing body for all of OpenAI Foundation and says each related organization is legally bound to pursue the organization's mission. Legal obligation and actual governance are different things, and that gap is where this trial will be won or lost ProPublica Nonprofit Explorer.

Musk's side has offered to direct any damages awarded to the nonprofit Ars Technica — a maneuver that reframes the suit from a billionaire's grievance into an act of institutional repair. OpenAI calls it harassment.

The jury will decide what the nonprofit was for, what it became, and who is responsible for the distance between the two.