Sam Altman’s testimony in the ongoing legal fight between OpenAI and Elon Musk has added a new layer to a dispute that is often framed as a battle over governance, contracts, and corporate control. But in court, Altman’s focus wasn’t only on what Musk did—or didn’t—legally. He described how Musk’s management approach allegedly reshaped the day-to-day culture of the research organization, arguing that it was fundamentally mismatched to the kind of environment a high-stakes AI lab needs.
According to Altman’s account, Musk pushed for an unusually aggressive internal ranking system—one that required senior leaders to evaluate researchers primarily through their accomplishments and then make sweeping personnel or performance changes based on those rankings. Altman said this process amounted to “huge damage” to the culture at OpenAI, using language that suggested not just disagreement with the method, but concern about the psychological and professional impact such a style can have inside a research institution.
The testimony, delivered as part of Musk’s lawsuit against OpenAI, centered on how Musk’s involvement affected morale and internal dynamics. Altman testified that Musk required OpenAI president Greg Brockman and former chief scientist Ilya Sutskever to rank researchers by their achievements and then “take a chainsaw through a bunch.” The phrase is striking because it implies more than routine performance management; it suggests a blunt, high-pressure intervention that could be experienced as destabilizing, even humiliating, particularly in a setting where researchers are expected to take long-term risks and pursue ideas that may not pay off immediately.
Altman acknowledged that this was consistent with the management style Musk is known for elsewhere. In other words, Altman did not portray Musk’s approach as random or unfamiliar. Instead, he argued that the same tactics that might work in other contexts—where speed, disruption, and rapid iteration are prized—were incompatible with running a research lab whose output depends on sustained intellectual effort, careful experimentation, and a culture that encourages deep thinking rather than constant fear of being cut.
That distinction matters, because research organizations don’t behave like factories. In a factory, you can measure output quickly and adjust processes with relatively direct cause-and-effect. In frontier AI research, the timeline is longer, the uncertainty is higher, and the “product” is often knowledge itself: new methods, new architectures, new training strategies, and sometimes entirely unexpected breakthroughs. When leadership applies a chainsaw-like evaluation model, the risk isn’t only that some people are judged unfairly—it’s that the entire system starts optimizing for the wrong incentives.
Altman’s testimony implicitly points to a core tension in AI organizations: how to balance accountability with psychological safety. Researchers need standards and expectations, but they also need room to explore. If the internal message becomes “prove yourself quickly or be removed,” then researchers may gravitate toward safer projects, incremental improvements, or work that is easier to defend in short time windows. That can reduce the lab’s ability to pursue the kinds of speculative ideas that often lead to major advances.
In court, Altman’s lawyer asked about the impact of Musk’s departure from OpenAI on morale. Altman’s response, as reported, was blunt: he said he didn’t think Musk understood how to run a good research lab. That line is important because it frames the issue not as a simple personality clash, but as a mismatch in operational philosophy. Running a research lab requires different leadership instincts than running a company built around rapid product cycles and aggressive restructuring.
To understand why Altman would characterize Musk’s approach as damaging, it helps to consider what “ranking researchers by accomplishments” can mean in practice. Accomplishments in research are notoriously difficult to compare across disciplines and time horizons. A researcher who publishes fewer papers might still be essential to a breakthrough through mentorship, infrastructure, or foundational work that enables others. Conversely, someone with a strong publication record might not be the best fit for a particular phase of a lab’s roadmap. Even when leadership tries to be fair, ranking systems can flatten nuance into a single score.
And once you introduce a high-stakes ranking system, you also introduce a second-order effect: researchers begin to manage themselves. They may spend more time on visibility—presenting results, aligning with leadership preferences, or choosing projects that look impressive on paper—rather than focusing on the hardest technical problems. In a field where progress can be nonlinear, that shift can quietly degrade the quality of work without any obvious sign on day one.
Altman’s “chainsaw” description suggests that the evaluation wasn’t merely a passive assessment. It implied action: sweeping changes following the rankings. That kind of intervention can create a climate where people feel replaceable, where collaboration becomes transactional, and where the cost of failure rises sharply. In research, failure is not an exception—it’s part of the process. If leadership treats failure as evidence of incompetence rather than as a normal step toward discovery, the lab’s culture can become brittle.
This is where Altman’s testimony becomes more than a personal critique. It becomes an argument about what kind of organizational behavior produces better scientific outcomes. A research lab tends to thrive when it can absorb setbacks, learn quickly, and keep talented people engaged through a sense of mission and belonging. If leadership repeatedly signals that people will be cut based on a narrow snapshot of accomplishments, the lab may lose not only individuals but also the collective willingness to take risks.
Altman’s account also highlights a recurring theme in high-profile tech disputes: the difference between external narratives and internal realities. Publicly, Musk and OpenAI have been associated with big ideas and ambitious timelines. But internally, the day-to-day experience of employees can diverge dramatically from the public story. Court testimony, by its nature, forces details into the open—details about meetings, decisions, and management practices that rarely make it into press coverage.
In this case, Altman’s testimony suggests that Musk’s involvement wasn’t only about strategic disagreements. It was about how power was exercised inside the organization. Requiring senior leaders to rank researchers and then make sweeping changes is a form of authority that can override existing norms. Even if the intent is to improve performance, the method can undermine trust. Researchers may start questioning whether leadership understands the work well enough to judge it, especially if the evaluation criteria are perceived as superficial or misaligned with the lab’s actual goals.
Altman’s statement that Musk “didn’t understand how to run a good research lab” can be read as a critique of both competence and empathy. Competence, because research labs require specialized judgment about what counts as progress. Empathy, because the human impact of abrupt, high-pressure management can be severe—particularly for scientists who have built their careers around autonomy, curiosity, and long-term thinking.
There is also a subtle but significant point embedded in Altman’s concession that Musk’s approach reflected a style Musk was known for. That concession matters because it suggests Altman wasn’t denying that Musk’s behavior was consistent with his reputation. Instead, Altman’s argument is that consistency doesn’t equal suitability. A management style can be effective in one environment and harmful in another. The question isn’t whether Musk is capable of decisive action; it’s whether that decisiveness translates into better outcomes for a research organization.
This is particularly relevant in AI, where the stakes are not only commercial but also societal. Frontier AI development involves intense scrutiny, regulatory uncertainty, and ethical concerns. Organizations operating in this space must maintain internal stability while moving quickly. If culture is damaged, the lab may struggle to retain talent, coordinate effectively, and sustain the focus needed to navigate both technical challenges and external pressure.
Altman’s testimony also arrives at a moment when the AI industry is increasingly aware of how organizational design affects model development. Many observers have noted that top AI labs are not just collections of engineers—they are ecosystems of researchers, data specialists, infrastructure teams, and product-minded strategists. The lab’s culture influences everything from how experiments are prioritized to how failures are handled to how knowledge is shared across teams. When leadership introduces fear-based incentives, it can distort the ecosystem.
That distortion can show up in ways that are hard to quantify. For example, researchers might become less willing to share partial results, fearing that incomplete work will be interpreted as lack of accomplishment. They might avoid proposing risky experiments because the downside feels too personal. Or they might spend more time defending their status rather than advancing the science. Over time, these behaviors can reduce the lab’s overall learning rate—even if individual outputs appear strong in the short term.
Altman’s “huge damage” framing suggests that, in his view, these effects were not hypothetical. He described the impact as real and significant enough to matter in court. That implies that the cultural harm was not limited to a few isolated incidents. Instead, it likely reflected a broader pattern of management interventions that changed how people felt about their work and their future inside the organization.
It’s also worth noting that the testimony references specific senior figures: Greg Brockman and Ilya Sutskever. Mentioning them is not just about name recognition. It signals that the alleged process wasn’t delegated to lower-level managers. It was something that reached the top of the organization, requiring leaders who are deeply embedded in the lab’s scientific and strategic direction to participate in the ranking and subsequent actions. When senior leaders are pulled into high-pressure evaluations, it can send a message throughout the organization that the lab’s internal norms are being overridden.
That message can be especially destabilizing in a research environment where credibility and trust are central. Scientists often rely on the judgment of peers and respected leaders. If the evaluation process is perceived as externally driven or misaligned with research realities, it can erode trust in leadership decisions. And once trust erodes, even good intentions can fail to produce the desired outcomes.
Altman’s testimony also underscores how legal disputes can become disputes about culture. Musk’s lawsuit against OpenAI is often discussed in terms of corporate structure and obligations. But when Altman describes management practices and their effects on morale, the case expands into a broader narrative about what kind of organization OpenAI was—and what it became under