Foreign media commentaries suggest that the focus of recent AI safety discussions is shifting from "how to reduce risks" to "who should set the rules." Following calls from several major labs to slow down advancements in frontier models, internal industry divisions over whether governments should intervene and whether companies can self-regulate have become increasingly public.
Large laboratories advocate slowing down progress.
Anthropic CEO Dario Amodei recently published an article advocating for slowing the pace of AI development to allow time for safety measures and to promote coordinated deployment between corporations and governments. OpenAI CEO Sam Altman and xAI founder Elon Musk have both expressed support for this direction.
However, supporting “greater emphasis on security” does not equate to supporting heavy regulation. Meta CEO Mark Zuckerberg stated that the company delayed the release of its AI model, Muse, for several months due to security and safety concerns. His remarks align more closely with another perspective: companies will incorporate security into their daily development processes due to market pressures and product responsibility, without waiting for unified regulations to come first.
The White House and Congress are taking a cautious stance.
The report noted that, despite calls from some leading companies for public-private collaboration, the U.S. government has not shown a clear willingness to take the lead. David Sacks, an AI official in the Trump administration, stated that AI regulation should primarily be entrusted to the companies developing the technology. The White House has also previously favored preventing individual states from implementing their own AI regulations.
There was also no rapid response at the congressional level. The article notes that after Amodei’s lengthy piece was published, U.S. House Speaker Mike Johnson expressed skepticism toward the notion of “catastrophic AI risks,” indicating that Washington is not actively pushing for the immediate establishment of a stronger regulatory framework.
Industry self-regulation raises new questions
Another line of interest is that companies such as OpenAI and Anthropic are reportedly pushing to establish an AI standards organization, which outsiders view as a potential industry self-regulatory body. If implemented, the AI industry would largely continue to self-regulate through voluntary commitments and standard-setting by corporations.
This also raises ongoing concerns: whether leading companies promoting "security rules" are also using them to solidify their own advantages. Critics argue that whether in the form of formal regulation or industry standards, if the thresholds are dominated by a few large companies, they may increase costs for newcomers and squeeze the competitive space for small and medium-sized enterprises.
China and overseas companies have raised objections.
The article also mentions that China has recently criticized certain U.S. companies for using the narrative of AI risks to disrupt global governance discussions. One reason is that Amodei explicitly stated in the article that the U.S. should widen its lead over the coming years by adjusting its commercial arrangements related to China.
Similar concerns are not limited to China. Aidan Gomez, CEO of the Canadian AI company Cohere, has also publicly stated that the debate has never been just about “whether to have guardrails,” but rather “who writes the rules, who gets to participate, and whose interests the rules protect.”
According to this review, AI safety is no longer just a technical issue but also a contest over industrial order and international competition. The real divide going forward may not be whether safety measures are needed, but rather who ultimately holds the power to set the rules.