Ilya Reveals the Inside Story of the OpenAI Struggle: Once Considered Merging with Anthropic and Preparing to Oust Altman for a Year

As the lawsuit by Musk against OpenAI continues to progress, co-founder and former chief scientist Ilya Sutskever appeared in court and submitted a 52-page memorandum, which fully reveals the basis for the board's decision to remove Sam Altman in 2023 for the first time. According to court records, Ilya stated that the reasons were not due to differences in business direction, but rather the collapse of trust at the executive level.

(OpenAI co-founder Ilya Sutskever has left, Sam Altman announces new Chief Scientist)

Ilya confessed that he had long had the idea of dismissing Sam Altman: preparation took more than a year.

The Information reported that the memorandum was requested to be written by directors Adam D’Angelo, Helen Toner, and Tasha McCauley, and was sent to the independent directors through self-destructing email (disappearing email) to prevent leaks.

Ilya emphasized in his testimony that the purpose of writing the document was to prompt the board to take action, and he believed that dismissing Altman was a reasonable decision.

He also acknowledged that he had considered ousting Altman at least a year before the incident, but postponed it due to an unfavorable board structure at the time. It wasn't until a “hostile majority against Altman” emerged that he proposed action.

Court documents quote a report from The Wall Street Journal: “Ilya has been waiting for the opportunity to replace Altman.” It now seems that this can also be confirmed in his testimony.

Recall movement trigger: From Altman's habitual lying to untrustworthiness

Regarding the underlying reasons, Ilya is in agreement with the previous positions of the two nonprofit oversight organizations, The Midas Project and The Tech Oversight Project, pointing out that Altman has a habit of lying and frequently makes misleading statements.

Ilya pointed out that activities such as inciting high-level relationships, creating chaos, ignoring security issues, lacking transparency in significant decisions, and being ambiguous about others' suggestions have made it impossible for the board to fulfill its supervisory responsibilities.

However, looking back at the recall process, Ilya believes it was still a bit rushed, but he emphasized that the fundamental reason for making such an extreme decision was that there was no longer any trust in the CEO himself regarding critical governance matters.

( Ten Thousand Character Report Reveals the Dark History of ChatGPT's Parent Company! Exposes Ultraman's Habitual Lying and Failure to Comply with Conflict of Interest, Musk Angrily Slams: Fraud )

Shocking revelation: OpenAI once discussed merging with Anthropic

The most shocking part is that Ilya also confirmed for the first time during this court appearance that after firing Altman, the board was not only looking for a new CEO but also actively reached out to its number one competitor, Anthropic, to discuss the possibility of a merger and acquisition.

According to him, Helen Toner informed the board of the proposal and subsequently had a call with Dario and Daniela, co-founders of Anthropic and former research and HR leaders at OpenAI.

Even though Ilya strongly opposed the plan at the time, most of the other directors supported it, and there was even speculation about Dario taking the position of CEO; however, the merger ultimately fell through due to execution difficulties.

Brockman Memo: Another Unpublished Key Document

During the questioning process, Ilya also admitted to having written another memorandum targeting OpenAI's former president and co-founder Greg Brockman, stating that the document is currently held by a lawyer.

The court has currently requested him to supplement his testimony and submit the document to establish a more complete context for the board's decision-making, examine whether Ilya has biases or distrust towards multiple executives, and clarify whether the decision to dismiss was due to institutional considerations rather than personal animosity. This document will undoubtedly become the focus of subsequent proceedings in the case.

Shareholding dispute and motive questioning: Ilya's conflict of interest

On the other hand, the legal team pressed Ilya on whether he would consider the loss of his stake value in OpenAI during the dismissal incident. Ilya expressed denial but refused to disclose the specific amount. The court listed this response as a potential “motive and bias” examination item, becoming part of the subsequent evidence collection.

The credibility of the evidence is questioned: mostly from second-hand information.

At the same time, Ilya also acknowledged that most of the content of his 52-page memo came from screenshots and accounts by OpenAI's former CTO Mira Murati, including Altman's lies to the board to cover up safety issues with GPT-4 Turbo.

He admitted that he did not personally verify with the parties involved and only wrote based on the information provided by Mira. Ilya also reflected on this in court, saying: “I later realized that first-hand evidence should take priority.”

Ilya discusses the context of impeachment: it's not a conflict of ideas, but a collapse of trust.

However, this testimony still provides the most authoritative legal document for OpenAI's internal struggle two years ago. With the release of the Brockman Memo and more internal emails, this conflict, known as “the most shocking power struggle in AI history,” is being redefined as a governance collapse disaster.

This article reveals Ilya's insider struggles at OpenAI: there was a desire to merge with Anthropic and preparations to oust Altman lasted for a year. First appeared on Chain News ABMedia.