OpenAI CEO Altman: Both Paths of AI Will Lead to Disaster
OpenAI CEO Sam Altman stated on X platform that there are two "directions to avoid" in AI development: the first is the complete loss of human control over the future, which he called "unacceptable," asserting that "we firmly stand with 'Team Humanity,' and AI must always serve humanity"; the second is excessive concentration of power—if an extremely powerful AI system is used by a single person or company to impose their worldview on everyone, "the outcome could be extremely dystopian." He emphasized that to avoid both risks, a "narrow middle path" must be taken, citing examples—one risk is a country gaining too much power, while another is a laboratory ultimately holding too much power; to avoid the first risk, it is essential to ensure that the development of alignment and safety technologies always outpaces the advancement of model capabilities.
This statement continues Altman's position, which he has reiterated in various public forums over the past month—on August 17, he stated on the "Relentless" podcast that "the real struggle now is: are we heading towards AI authoritarianism or freedom," claiming that "every time humanity sacrifices freedom for safety, it is a net loss in the long run"; on August 23, in an interview with podcast host David Senra, he further stated, "The right approach is to allow people to deeply control the future, with society and models evolving together."
Notably, multiple reports have pointed out that Altman's stance has tension with OpenAI's own business practices—the company's flagship model still operates in a closed-source, proprietary manner; in contrast, Anthropic is one of the few leading labs that has neither signed nor publicly supported the "Open Weights and U.S. AI Leadership" open letter, instead emphasizing stronger safety protections for increasingly capable models. This divergence has previously been interpreted as one of the signals of Altman "implicitly criticizing" Anthropic's conservative stance.
In market mechanisms, Altman's repeated emphasis on the risk of "power concentration" comes during a sensitive window as OpenAI negotiates cross-laboratory safety coordination mechanisms with Anthropic, xAI, and Google DeepMind, and as OpenAI prepares for its IPO; positioning "to avoid power concentration in a single laboratory" as a public stance objectively helps OpenAI position itself as the "proponent of decentralization and democratization" in this safety narrative competition, distinguishing itself from critics who accuse it of "building a regulatory moat in the name of safety." This is also part of the discourse competition among leading AI labs regarding "whose solution is more trustworthy" before regulatory frameworks are established.
ABAB AI Insight
In recent months, Altman has repeatedly emphasized the risk of "power concentration" in various podcasts and public forums, essentially reinforcing his previous binary framework of "AI authoritarianism vs. freedom"—this rhetoric can be traced back to his blog post "The Gentle Singularity" published in July, where he proposed the principle that "superintelligence should not be overly concentrated in any individual, company, or country." This recent post on X is a continuation and further specification of this idea.
From the perspective of resource allocation, while Altman publicly warns of the risk of "power concentration," OpenAI, which he leads, still maintains a closed-source business model and is advancing its IPO process (according to previous disclosures, the company has secretly submitted an IPO application with an estimated valuation of about $852 billion)—this combination of "calling for power decentralization while continuing to expand capital and market influence" has been viewed by some commentators as a discourse strategy: by occupying the moral high ground of "democratizing AI," he seeks to gain a public opinion advantage in the safety narrative competition against competitors like Anthropic, without necessarily corresponding to specific open-source or decentralization actions.
This paradox of "warning against concentration risks while simultaneously moving towards concentration" is similar to the public relations strategies of several industry giants under antitrust pressure—large tech companies often advocate for principles like "open ecosystems" and "data portability rights" when facing regulatory scrutiny, even as their business models still heavily rely on platform monopolies; in terms of industry positioning, leading AI labs are currently engaged in a three-way game involving safety narratives, regulatory frameworks, and capital market valuations, with the discourse power around "power concentration" itself becoming an important bargaining chip influencing regulatory directions and investor confidence.
Essentially, this is a discourse power precondition regarding the future shape of regulatory frameworks—if "power concentrated in a single laboratory" is widely accepted as a core risk, the natural direction of regulatory policy will be to mandate leading labs to open model weights and introduce external governance mechanisms, which would favor companies like OpenAI that have a richer open-source product line in seizing the regulatory high ground, while being relatively unfavorable to Anthropic, which insists on a completely closed-source model and emphasizes "safety over openness"; mechanism-wise, whoever can package their business model as a "risk antidote" before the regulatory framework is established will have greater initiative in the final formulation of regulatory details.
ABAB News · Cognitive Law
- The loudest calls for "anti-concentration" are often a prelude to concentration on their own side.
- Before the regulatory framework is finalized, whoever tells the best story will gain the authority to set the rules.
- The open-source vs. closed-source debate is fundamentally about whose safety narrative can better persuade regulators.