OpenAI CEO Sam Altman has reiterated his warnings about the potential dangers of artificial intelligence, outlining two critical scenarios where AI progress could severely harm humanity. He is also lending his support to a global initiative urging companies to temper the pace of development in advanced AI models.
Altman stated on X, formerly Twitter, that the two primary risks to avoid are "losing control of the future to AI" and the "concentration of power" where a single entity could impose its worldview on others, leading to a "dystopian" outcome. He emphasized that OpenAI is "unapologetically on Team Humanity."
Altman and Elon Musk, who have previously been in legal opposition, have joined forces to back a statement by Anthropic CEO Dario Amodei. This statement calls for global companies to decelerate their work on leading AI models. The concerns were amplified by Anthropic researcher Jacob Coxon, who recently resigned, citing fears that companies like Anthropic and OpenAI are "gambling with our lives" and that AI could pose an existential threat by the end of the decade.
Amidst a global race for AI supremacy, the United States and China are at the forefront. US tech leaders have voiced concerns that advanced AI could empower China with superior surveillance, military capabilities, and cybersecurity. However, China has dismissed these warnings. Guo Jiakun, a spokesperson for China's Ministry of Foreign Affairs, characterized such calls for slowing AI advancement as "fear mongering" and indicative of "Cold War" tactics, a sentiment echoed by state-backed media.
President Xi Jinping has indicated China's intent to lead in fostering global AI collaboration among developing nations. In contrast, President Trump has downplayed these concerns, asserting that such catastrophic predictions are unlikely and that the U.S. is currently leading the AI race against China.
The broader market has reacted to these discussions, with tech stocks worldwide experiencing declines. Investors are reportedly concerned about the substantial investments made in AI development potentially yielding reduced profits if progress is significantly slowed. Major companies like Nvidia, AMD, SpaceX, and Amazon saw their stock values drop, as did European and Asian tech firms such as ASML, Infineon, Siemens Energy, SoftBank, TSMC, and SK Hynix.
Altman has also proposed specific measures for implementing global AI safeguards. These include federal regulations and the establishment of independent auditors. He stressed that all leading AI labs must commit to these standards, stating, "We think shared standards for misalignment, monitoring, and safety will lead to better outcomes." While acknowledging that progress will continue, Altman advocated for a deliberate, slower pace, stating, "But it should be slower than it otherwise could be."
Government intervention is seen as crucial for securing international cooperation, though this presents a significant challenge. OpenAI had been preparing for a stock market debut, but Altman announced on Saturday that OpenAI would not go public in 2026 due to safety considerations.
Amodei's essay detailed risks associated with recursive self-improvement (RSI), where AI systems gain the ability to advance independently. He noted a recent acceleration in AI progress, largely driven by AI's capacity to develop subsequent generations of AI. Amodei suggested a three-pronged approach that would protect commercial interests and the U.S.'s lead in AI. This plan involves granting external evaluators "employee-like access" to AI models, establishing common safety standards among AI labs, and limiting the speed of unchecked AI advancement.
Amodei recognized the difficulty of achieving global cooperation, citing strong incentives for nations to gain a competitive and military advantage. "I don’t know if it’s possible, but we should try," he remarked in an interview with CBS News' "Sunday Morning."