More from this creator
Other episodes by Kitty Cat.
Transcript
The full episode, in writing.
Elon Musk’s lawsuit has brought OpenAI’s safety record and internal governance into sharp focus. On May 7, 2026, a federal court in Oakland, California, heard testimony from Rosie Campbell, a former OpenAI employee and board member, about the company’s evolving approach to artificial intelligence safety. Campbell joined OpenAI’s AGI readiness team in 2021 and left in 2024 after her team, along with the Super Alignment team, was disbanded. She described a significant shift within the company from a research-driven culture focused on artificial general intelligence safety, to one increasingly centered on product development and commercial deployment.
Campbell testified about an incident where Microsoft, a prominent OpenAI partner, deployed a version of OpenAI’s GPT-4 model in India through its Bing search engine. This deployment occurred before the model had been evaluated by OpenAI’s Deployment Safety Board, a group responsible for reviewing safety risks before product releases. Campbell stated that while the specific model did not present a major risk, the precedent of bypassing established safety review was concerning. She emphasized the need for strong and consistently followed safety processes as AI technologies become more powerful, arguing that reliable safety frameworks are essential for responsible development and deployment.
Under cross-examination, Campbell acknowledged that significant funding is necessary to achieve OpenAI's goal of building advanced AI systems, but she expressed skepticism that financial imperatives should override foundational safety principles. When questioned about comparative safety practices, Campbell admitted that in her speculative opinion, OpenAI’s current approach to safety is superior to that at xAI, the artificial intelligence company founded by Elon Musk. xAI was acquired by SpaceX earlier in 2026, highlighting Musk’s ongoing involvement in the AI sector.
OpenAI has responded to scrutiny by releasing public evaluations of its AI models and sharing its safety framework. However, the company declined to comment specifically on its current approach to AGI alignment during the court proceedings. The current head of preparedness at OpenAI is Dylan Scandinaro, who was hired from Anthropic in February. Sam Altman, OpenAI’s CEO at the time, said that Scandinaro’s hire would let him “sleep better tonight,” suggesting the role’s importance in the organization’s safety structure.
The deployment of GPT-4 in India was a contributing factor to the OpenAI board’s brief decision to fire CEO Sam Altman in 2023. This decision followed complaints from key employees, including then-chief scientist Ilya Sutskever and then-CTO Mira Murati, about Altman’s management style, particularly regarding insufficient confrontation of conflicts and failures in transparency. Tasha McCauley, a board member at the time, testified about concerns that Altman was not sufficiently forthcoming with the board, undermining the unusual structure in which a non-profit board was meant to oversee a for-profit subsidiary.
McCauley also described a pattern of Altman misleading the board. For example, Altman lied to a board member about McCauley’s intention to remove Helen Toner, another board member who had published a white paper implicitly criticizing OpenAI’s safety policy. Altman further failed to inform the board about the public launch of ChatGPT and did not disclose potential conflicts of interest. McCauley stated, “Our primary way to do that [oversee the for-profit subsidiary] was being called into question. We did not have a high degree of confidence at all to trust that the information being conveyed to us allowed us to make decisions in an informed way.”
The board’s decision to remove Altman coincided with a tender offer to OpenAI employees. As staff began to support Altman and Microsoft intervened to restore his position, the board members opposed to Altman ultimately stepped down, reversing their previous decision. This sequence of events highlighted the limited ability of the non-profit board to influence the trajectory of OpenAI’s for-profit operations. Elon Musk’s lawsuit argues that this breakdown in internal governance constitutes a breach of the organization’s original mission, which was to ensure the benefits of artificial general intelligence are shared by all of humanity.
David Schizer, former dean of Columbia Law School, appeared as an expert witness for Musk’s legal team. Schizer reiterated the concerns about OpenAI’s commitment to safety, stating, “OpenAI has emphasized that a key part of its mission is safety and they are going to prioritize safety over profits. Part of that is taking safety rules seriously, if something needs to be subject to safety review, it needs to happen. What matters is the process issue.” Schizer’s testimony underscored the principle that robust internal processes are critical for upholding a safety-first approach in the development of advanced AI systems.
The issues at OpenAI extend beyond a single company and reflect broader challenges in the governance of profit-driven AI research. McCauley testified that the failures of OpenAI’s internal oversight suggest a need for stronger government regulation of advanced AI technologies. She argued that leaving decisions with far-reaching societal implications to a single CEO, especially when the public good is at stake, is “very suboptimal.”
OpenAI’s safety framework includes public documentation of model evaluations, but the details of its current AGI alignment strategy remain undisclosed. The company’s for-profit subsidiary structure, combined with its rapid growth into one of the world’s largest private companies, has fueled debate over whether the original non-profit mission has been compromised. The company’s trajectory from research-focused non-profit to commercial powerhouse has been marked by internal disputes, shifting priorities, and external scrutiny.
The Super Alignment team, which was tasked with ensuring AI systems could be reliably controlled even as they approached human-level intelligence, was one of the groups disbanded during the organizational shift. Rosie Campbell’s AGI readiness team, created to anticipate and address potential risks associated with artificial general intelligence, was also eliminated in 2024. These structural changes signaled a move away from proactive safety research in favor of accelerating product development.
Microsoft’s partnership with OpenAI has grown increasingly influential, particularly in the distribution and deployment of advanced language models like GPT-4. The incident in India, where a version of GPT-4 was released before internal safety review, illustrated the tension between rapid market expansion and the enforcement of safety protocols. This case added to the arguments that OpenAI’s priorities have become misaligned with its founding principles.
Helen Toner, the board member whose removal was concealed from the board by Sam Altman, had published research that implicitly criticized OpenAI’s safety approach. Her white paper, discussed in McCauley’s testimony, became a flashpoint for the organization’s internal debate over transparency and dissent. The board’s limited power to enforce accountability ultimately contributed to the leadership crisis in 2023.
The acquisition of xAI by SpaceX in 2026 brought another Musk-founded AI company into the spotlight. This move further entwined the interests of private space and AI ventures, raising questions about the future direction and oversight of powerful new technologies. Dylan Scandinaro, brought in from Anthropic, now leads OpenAI’s preparedness efforts, reflecting a trend of hiring from competing labs with shared safety concerns.
TechCrunch Disrupt 2026 is set to bring together over 10,000 founders, investors, and technology leaders in San Francisco from October 13 to 15, with more than 250 sessions scheduled. This event highlights the growing scale and commercial influence of the AI sector, marking a stark contrast to OpenAI’s origins as a mission-driven research collective.
The breakdown in OpenAI’s internal oversight, Microsoft’s role in model deployment, and the disbanding of safety teams have all factored into Elon Musk’s legal strategy to challenge OpenAI’s transformation from a non-profit research lab into a major for-profit enterprise.