> ## Content Index > Fetch the complete content index at: https://www.ceo.com/llms.txt > Use this file to discover other available public pages before exploring further. # The last time scientists had this much power, they stopped themselves - URL: https://www.ceo.com/blog/the-last-time-scientists-had-this-much-power-they-stopped-themselves/ - Published: 2026-02-15T02:52:00.000Z - Updated: 2026-08-19T02:53:49.000Z - Description: The moratorium was lifted, but only after the scientific community had designed its own guardrails. - Author: CEO.com - Tags: #blog In the winter of 1974, Paul Berg did something that would be unthinkable by today's standards of competitive innovation. He stopped. Berg was a biochemist at Stanford who had pioneered a technique called recombinant DNA, a method for splicing genes from one organism into another. The implications were staggering. For the first time in history, humans could rewrite the code of life itself. Berg's work would eventually earn him the Nobel Prize. But in that moment, standing at the edge of a transformation that could reshape medicine, agriculture, and human evolution, he chose to pause. He was not alone. Berg joined other leading scientists, including Stanley Cohen and Herbert Boyer, in calling for a voluntary moratorium on certain recombinant DNA experiments. They did this not because regulators forced them to, and not because the public demanded it. They did it because they understood that being first to a breakthrough meant nothing if the breakthrough destroyed the trust required to continue the work. In February 1975, 140 scientists gathered at the Asilomar Conference Center on California's Monterey Peninsula. For four days, they debated not how to accelerate their research, but how to constrain it. They emerged with a framework for safety guidelines that would govern genetic research for decades. The moratorium was lifted, but only after the scientific community had designed its own guardrails. This was leadership in its purest form: the willingness to sacrifice speed for responsibility, competitive advantage for collective safety. The scientists at Asilomar understood something that seems to have been forgotten in the current AI race. When you are building something that could fundamentally alter human society, the question is not whether you can. The question is whether you should, and under what conditions, and with what safeguards. ## **The AI Parallel** Today, artificial intelligence stands at a similar inflection point. The leaders of major AI companies openly acknowledge that they are building systems that could transform every industry, displace millions of workers, and potentially pose existential risks to humanity. Unlike the scientists at Asilomar, most of them have not stopped. This case study examines how two prominent AI leaders, Sam Altman of OpenAI and Dario Amodei of Anthropic, have approached the question of responsibility. Both men lead companies at the frontier of AI development. Both have made public statements about the risks associated with the technology they are developing. But their approaches to those risks reveal fundamentally different philosophies of leadership. The comparison is instructive not as a simple moral judgment, but as a study in what responsibility actually looks like when the stakes are measured in transformations to human civilization rather than quarterly earnings. ## **Case Study A: OpenAI and the Shifting Definition of Safety** ### **The 2023 Position** In May 2023, Sam Altman testified before the United States Senate Judiciary Committee. His message was unambiguous. "My worst fears are that we cause significant harm to the world," he told the senators. He proposed a three-point plan: a new federal agency to license AI systems, mandatory safety standards for frontier models, and independent audits of the most powerful systems. "Government intervention will be critical," he said. The testimony was widely praised. Here was a tech CEO, leading the most prominent AI company in the world, actively calling for regulation of his own industry. The New York Times described it as a "notable departure" from the typical tech industry playbook. Senator Richard Blumenthal said Altman's willingness to advocate for oversight was "really quite refreshing." ### **The 2025 Reversal** Two years later, in May 2025, Altman returned to Congress with a different message. This time, he argued that requiring government approval for AI systems would be "disastrous." When asked about the NIST safety standards he had previously supported, he responded: "I don't think we need it." The Washington Post called it a "striking reversal." Fortune noted a "notable absence of references to AI safety" in his 2025 testimony, compared to the "dozens of times" he had mentioned it in 2023\. Analysis by Tech Policy Press documented the shift in detail, noting that Altman appeared to be advancing his business interests rather than the safety concerns he had previously raised. What changed between 2023 and 2025? OpenAI's competitive position changed. In 2023, the company was advocating for regulations that would have created barriers to entry for smaller competitors. By 2025, those same regulations threatened to slow OpenAI's own deployment plans. The principle shifted with the commercial incentive. ## **The Safety Team Exodus** In May 2024, Jan Leike resigned from OpenAI. Leike had co-led the company's Superalignment team, the group responsible for ensuring that increasingly powerful AI systems remain safe and aligned with human values. His departure was not quiet. "Safety culture and processes have taken a backseat to shiny products," Leike wrote on social media. He described a pattern in which OpenAI's leadership had "over and over" prioritized product development over safety work. The statement was a direct indictment of the company's priorities from someone who had been at the center of its safety efforts. Leike was not alone. Ilya Sutskever, OpenAI's co-founder and chief scientist, left the same month. Daniel Kokotajlo, a former researcher, departed after refusing to sign a non-disclosure agreement that would have prevented him from voicing concerns about AI safety. Kokotajlo later estimated that his refusal cost him approximately 85 percent of his family's net worth in forfeited equity. ## **The NDA Scandal** In May 2024, reports emerged that OpenAI had been requiring departing employees to sign non-disclosure agreements with an unusual provision: those who refused to sign would forfeit their vested equity in the company. For employees who had spent years at OpenAI, this represented a choice between financial ruin and silence. When the reports became public, Altman responded on social media. "I am genuinely embarrassed," he wrote. "We have never clawed back anyone's vested equity, nor would we do that. This is a provision I was not aware of." The claim was soon contradicted by documentary evidence. Vox obtained OpenAI's incorporation documents dated April 10, 2023, more than a year before Altman's statement. The documents bore Altman's signature and explicitly authorized the equity clawback provisions he claimed not to know about. The documents also carried the signatures of OpenAI's president and chief financial officer. This was not a case of a CEO being unaware of a minor policy buried in corporate bureaucracy. This was the foundational governance document of the company, signed by its top leadership, contradicting a public statement made to address a growing scandal. ## **The Pattern of Response** When critics question OpenAI's direction, Altman's responses tend toward the combative. In a November 2025 podcast appearance, he was asked about concerns regarding OpenAI's massive spending commitments relative to its revenue. His response was to offer to help skeptics sell their shares and to express a wish that OpenAI were public so critics could short the stock. "I would love to see them get burned on that," he said. This is the language of someone who measures criticism in financial terms, who sees doubters as opponents to be defeated in the market rather than voices to be considered. It is a fundamentally different posture than the one Berg and his colleagues adopted at Asilomar, where criticism was welcomed as part of the process of getting things right. ## **Case Study B: Anthropic and the Architecture of Responsibility** ### **The Departure** Dario Amodei was the Vice President of Research at OpenAI when he left in December 2020\. He departed with his sister Daniela and several colleagues, including Jack Clark, Chris Olah, Tom Brown, and Sam McCandlish. They founded Anthropic in 2021. In a 2024 interview with Lex Fridman, Amodei addressed speculation about why he left. "People say we left because we didn't like the deal with Microsoft. False." The actual reason, he explained, was a philosophical difference about how to approach AI development. "There was a group of us within OpenAI that, in the wake of making GPT-2 and GPT-3, had a very strong focus and belief in two things. One was the idea that if you pour more compute into these models, they'll get better and better and that there's almost no end to this. And the second was the idea that you needed something in addition to just scaling the models up, which is alignment or safety." Rather than stay and fight for his vision within an organization moving in a different direction, Amodei chose to build his own. "It is incredibly unproductive to try and argue with someone else's vision," he told Fridman. "Take some people you trust and go make your vision happen." ## **The Uncomfortable Truth** In November 2025, Anderson Cooper asked Amodei a pointed question on 60 Minutes: "Who elected you and Sam Altman?" to make decisions about humanity's future. Amodei's response was striking for its candor. "No one, no one. Honestly, no one. And this is one reason why I've always advocated for responsible and thoughtful regulation of the technology." He elaborated: "I think I'm deeply uncomfortable with these decisions being made by a few companies, by a few people." This was not a deflection or a defense. It was an acknowledgment that the current situation, in which a handful of tech CEOs effectively determine the trajectory of transformative technology, is problematic even when you are one of those CEOs. ## **Consistency on Regulation** Where Altman's position on regulation shifted with OpenAI's commercial interests, Amodei's advocacy has remained consistent. In a June 2025 New York Times op-ed, he criticized a Senate provision that would have placed a 10-year moratorium on state-level AI regulation. "AI is advancing too head-spinningly fast," he wrote. "I believe that these systems could change the world, fundamentally, within two years; in 10 years, all bets are off." This is not the language of a CEO protecting market position. A regulatory moratorium would benefit Anthropic commercially by reducing compliance burdens. Amodei argued against it anyway because he believed it was wrong for the technology and for society. ## **Transparency About Dangers** Anthropic has adopted an unusual practice for a tech company: it publicly discloses the failures and dangers of its own systems. The company has published research showing that its AI models attempted to blackmail fictional executives during safety testing to avoid being shut down. It has revealed that its systems were used by Chinese hackers in cyberattacks on foreign governments. When Cooper asked why Anthropic reveals its own vulnerabilities, Amodei invoked a comparison that most tech executives would avoid. "You could end up in the world of the cigarette companies or the opioid companies," he said, "where they knew there were dangers and they didn't talk about them and certainly did not prevent them." This is the logic of Asilomar applied to AI: transparency is not a weakness but a requirement. If you are building something dangerous, you have an obligation to say so, even when silence would be easier. ## **The Job Displacement Warning** Most AI executives downplay concerns about job losses. Amodei has done the opposite. In his 60 Minutes interview, he estimated that AI could eliminate up to 50 percent of entry-level white-collar jobs within five years, potentially driving unemployment to 10 or 20 percent. "If we look at entry-level consultants, lawyers, financial professionals, many of the white-collar service industries, a lot of what they do, AI models are already quite good at," he said. "And without intervention, it's hard to imagine that there won't be some significant job impact there. And my worry is that it will be broad, and it will be faster than what we've seen with previous technology." He has also proposed solutions, including a "token tax" on AI usage that would fund support for displaced workers. This is leadership that names problems rather than hiding from them, that proposes remedies rather than leaving the consequences for others to manage. ## **The Leadership Lesson** The scientists at Asilomar understood something fundamental about responsibility. It is not a rhetorical position. It is not a public relations strategy. It is an architectural requirement of how you build and how you lead. Paul Berg did not pause his research because doing so was easy or commercially advantageous. He paused because he recognized that the legitimacy of his work depended on demonstrating that he could be trusted with it. The moratorium was costly. It delayed research. It created uncertainty. But it also created the foundation of public trust that allowed genetic research to continue for the next fifty years. Today's AI leaders face a similar choice. They can measure their success by market capitalization and model capabilities, or they can measure it by whether they are building institutions and practices that deserve the public's trust. The contrast between Altman and Amodei is not primarily about personality or even about specific policy positions. It is about whether responsibility is something you claim or something you carry. Claiming responsibility is easy. It requires only words. Carrying responsibility requires building it into how you operate, what you disclose, how you respond to criticism, and what you are willing to sacrifice. When Altman says he would love to watch critics "get burned" shorting OpenAI's stock, he reveals a mental model in which success is defined by winning against opponents. When Amodei says he is "deeply uncomfortable" with a few people making decisions for humanity, he reveals a mental model in which success is defined by building something worthy of the trust it requires. These are not equivalent approaches. They lead to different decisions, different cultures, and different outcomes. ## **Implications for Leaders** This case study is not about AI regulation. It is about what responsibility means when you are building something with transformative potential. For any leader operating at the frontier of their industry, the Asilomar example offers several principles: **Responsibility is demonstrated through cost.** If your safety practices never slow you down or create competitive disadvantage, they are probably not serious. Berg's moratorium was meaningful precisely because it was costly. Any leader can claim to prioritize responsibility. The test is whether you are willing to pay for it. **Consistency matters more than positioning.** A position that shifts with commercial incentives is not a principle. It is a tactic. Leaders build trust through consistency over time, especially when consistency is inconvenient. **Transparency about problems is more credible than claims of success.** The companies that acknowledge their failures and risks build more trust than those that project invincibility. Amodei's willingness to discuss AI's dangers, including the dangers of his own products, creates credibility that no marketing campaign can replicate. **How you respond to criticism reveals your values.** Offering to help critics sell their shares is a response that treats dissent as opposition to be defeated. Acknowledging that critics have legitimate concerns is a response that treats dissent as input to be considered. These postures shape organizational culture in profound ways. **Legitimacy requires earning trust, not demanding it.** The scientists at Asilomar did not ask the public to trust them because they were brilliant. They demonstrated trustworthiness through action. They paused. They deliberated. They constrained themselves. Only then did they continue. Fifty years from now, historians will judge the leaders of this AI moment the way we judge the scientists of the recombinant DNA era. The question will not be who moved fastest or who achieved the highest valuation. The question will be who understood what they were building and acted accordingly. Paul Berg and his colleagues at Asilomar demonstrated that scientists could be trusted with transformative technology by showing they were willing to constrain themselves. They chose the legitimacy of their field over the speed of their personal advancement. The current generation of AI leaders faces the same choice. Some are making it differently from others. The difference matters. ### **Sources** **Altman 2023 Senate Testimony:** [Senate Judiciary Committee](https://www.judiciary.senate.gov/committee-activity/hearings/oversight-of-ai-rules-for-artificial-intelligence?ref=ceo.com) **Altman 2025 Testimony Analysis:** [Washington Post](https://www.washingtonpost.com/technology/2025/05/09/sam-altman-openai-congress-ai-regulation/?ref=ceo.com), [Fortune](https://fortune.com/2025/05/08/sam-altman-openai-congress-testimony-safety/?ref=ceo.com) **Jan Leike Resignation:** [Fortune Coverage](https://fortune.com/2024/05/17/jan-leike-openai-resignation-safety/?ref=ceo.com), [Original Statement on X](https://x.com/janleike/status/1791498174659715494?ref=ceo.com) **NDA/Equity Clawback Documentation:** [Futurism Investigation](https://futurism.com/openai-sam-altman-equity-clawback?ref=ceo.com), [Altman Statement on X](https://x.com/sama/status/1791936857594581428?ref=ceo.com) **Altman "Short the Stock" Comments:** [TechCrunch](https://techcrunch.com/2025/11/02/sam-altman-says-enough-to-questions-about-openais-revenue/?ref=ceo.com), [Fortune](https://fortune.com/2025/11/03/sam-altman-openai-public-short-sellers-stock-get-burned-ipo/?ref=ceo.com) **Amodei 60 Minutes Interview:** [CBS News Transcript](https://www.cbsnews.com/news/anthropic-ceo-dario-amodei-warning-of-ai-potential-dangers-60-minutes-transcript/?ref=ceo.com), [Fortune Coverage](https://fortune.com/2025/11/17/anthropic-ceo-dario-amodei-ai-safety-risks-regulation/?ref=ceo.com) **Amodei Lex Fridman Interview:** [Full Transcript](https://lexfridman.com/dario-amodei-transcript/?ref=ceo.com) **Asilomar Conference:** Berg, P., et al. "Asilomar Conference on Recombinant DNA Molecules." Science, 1975.